8 .. ocv:class:: DataType
10 Template "trait" class for OpenCV primitive data types. A primitive OpenCV data type is one of ``unsigned char``, ``bool``, ``signed char``, ``unsigned short``, ``signed short``, ``int``, ``float``, ``double``, or a tuple of values of one of these types, where all the values in the tuple have the same type. Any primitive type from the list can be defined by an identifier in the form ``CV_<bit-depth>{U|S|F}C(<number_of_channels>)``, for example: ``uchar`` ~ ``CV_8UC1``, 3-element floating-point tuple ~ ``CV_32FC3``, and so on. A universal OpenCV structure that is able to store a single instance of such a primitive data type is
11 :ocv:class:`Vec`. Multiple instances of such a type can be stored in a ``std::vector``, ``Mat``, ``Mat_``, ``SparseMat``, ``SparseMat_``, or any other container that is able to store ``Vec`` instances.
13 The ``DataType`` class is basically used to provide a description of such primitive data types without adding any fields or methods to the corresponding classes (and it is actually impossible to add anything to primitive C/C++ data types). This technique is known in C++ as class traits. It is not ``DataType`` itself that is used but its specialized versions, such as: ::
15 template<> class DataType<uchar>
17 typedef uchar value_type;
18 typedef int work_type;
19 typedef uchar channel_type;
20 enum { channel_type = CV_8U, channels = 1, fmt='u', type = CV_8U };
23 template<typename _Tp> DataType<std::complex<_Tp> >
25 typedef std::complex<_Tp> value_type;
26 typedef std::complex<_Tp> work_type;
27 typedef _Tp channel_type;
28 // DataDepth is another helper trait class
29 enum { depth = DataDepth<_Tp>::value, channels=2,
30 fmt=(channels-1)*256+DataDepth<_Tp>::fmt,
31 type=CV_MAKETYPE(depth, channels) };
35 The main purpose of this class is to convert compilation-time type information to an OpenCV-compatible data type identifier, for example: ::
37 // allocates a 30x40 floating-point matrix
38 Mat A(30, 40, DataType<float>::type);
40 Mat B = Mat_<std::complex<double> >(3, 3);
41 // the statement below will print 6, 2 /*, that is depth == CV_64F, channels == 2 */
42 cout << B.depth() << ", " << B.channels() << endl;
45 So, such traits are used to tell OpenCV which data type you are working with, even if such a type is not native to OpenCV. For example, the matrix ``B`` initialization above is compiled because OpenCV defines the proper specialized template class ``DataType<complex<_Tp> >`` . This mechanism is also useful (and used in OpenCV this way) for generic algorithms implementations.
52 Template class for 2D points specified by its coordinates
55 An instance of the class is interchangeable with C structures, ``CvPoint`` and ``CvPoint2D32f`` . There is also a cast operator to convert point coordinates to the specified type. The conversion from floating-point coordinates to integer coordinates is done by rounding. Commonly, the conversion uses this
56 operation for each of the coordinates. Besides the class members listed in the declaration above, the following operations on points are implemented: ::
65 double value = norm(pt); // L2 norm
69 For your convenience, the following type aliases are defined: ::
71 typedef Point_<int> Point2i;
72 typedef Point2i Point;
73 typedef Point_<float> Point2f;
74 typedef Point_<double> Point2d;
78 Point2f a(0.3f, 0.f), b(0.f, 0.4f);
79 Point pt = (a + b)*10.f;
80 cout << pt.x << ", " << pt.y << endl;
85 .. ocv:class:: Point3_
87 Template class for 3D points specified by its coordinates
91 An instance of the class is interchangeable with the C structure ``CvPoint2D32f`` . Similarly to ``Point_`` , the coordinates of 3D points can be converted to another type. The vector arithmetic and comparison operations are also supported.
93 The following ``Point3_<>`` aliases are available: ::
95 typedef Point3_<int> Point3i;
96 typedef Point3_<float> Point3f;
97 typedef Point3_<double> Point3d;
103 Template class for specifying the size of an image or rectangle. The class includes two members called ``width`` and ``height``. The structure can be converted to and from the old OpenCV structures
104 ``CvSize`` and ``CvSize2D32f`` . The same set of arithmetic and comparison operations as for ``Point_`` is available.
106 OpenCV defines the following ``Size_<>`` aliases: ::
108 typedef Size_<int> Size2i;
110 typedef Size_<float> Size2f;
116 Template class for 2D rectangles, described by the following parameters:
118 * Coordinates of the top-left corner. This is a default interpretation of ``Rect_::x`` and ``Rect_::y`` in OpenCV. Though, in your algorithms you may count ``x`` and ``y`` from the bottom-left corner.
119 * Rectangle width and height.
121 OpenCV typically assumes that the top and left boundary of the rectangle are inclusive, while the right and bottom boundaries are not. For example, the method ``Rect_::contains`` returns ``true`` if
125 x \leq pt.x < x+width,
126 y \leq pt.y < y+height
128 Virtually every loop over an image
129 ROI in OpenCV (where ROI is specified by ``Rect_<int>`` ) is implemented as: ::
131 for(int y = roi.y; y < roi.y + rect.height; y++)
132 for(int x = roi.x; x < roi.x + rect.width; x++)
138 In addition to the class members, the following operations on rectangles are implemented:
141 :math:`\texttt{rect} = \texttt{rect} \pm \texttt{point}` (shifting a rectangle by a certain offset)
144 :math:`\texttt{rect} = \texttt{rect} \pm \texttt{size}` (expanding or shrinking a rectangle by a certain amount)
146 * ``rect += point, rect -= point, rect += size, rect -= size`` (augmenting operations)
148 * ``rect = rect1 & rect2`` (rectangle intersection)
150 * ``rect = rect1 | rect2`` (minimum area rectangle containing ``rect2`` and ``rect3`` )
152 * ``rect &= rect1, rect |= rect1`` (and the corresponding augmenting operations)
154 * ``rect == rect1, rect != rect1`` (rectangle comparison)
156 This is an example how the partial ordering on rectangles can be established (rect1
157 :math:`\subseteq` rect2): ::
159 template<typename _Tp> inline bool
160 operator <= (const Rect_<_Tp>& r1, const Rect_<_Tp>& r2)
162 return (r1 & r2) == r1;
166 For your convenience, the ``Rect_<>`` alias is available: ::
168 typedef Rect_<int> Rect;
172 .. ocv:class:: RotatedRect
174 The class represents rotated (i.e. not up-right) rectangles on a plane. Each rectangle is specified by the center point (mass center), length of each side (represented by cv::Size2f structure) and the rotation angle in degrees.
176 .. ocv:function:: RotatedRect::RotatedRect()
177 .. ocv:function:: RotatedRect::RotatedRect(const Point2f& center, const Size2f& size, float angle)
178 .. ocv:function:: RotatedRect::RotatedRect(const CvBox2D& box)
180 :param center: The rectangle mass center.
181 :param size: Width and height of the rectangle.
182 :param angle: The rotation angle in a clockwise direction. When the angle is 0, 90, 180, 270 etc., the rectangle becomes an up-right rectangle.
183 :param box: The rotated rectangle parameters as the obsolete CvBox2D structure.
185 .. ocv:function:: void RotatedRect::points( Point2f pts[] ) const
186 .. ocv:function:: Rect RotatedRect::boundingRect() const
187 .. ocv:function:: RotatedRect::operator CvBox2D() const
189 :param pts: The points array for storing rectangle vertices.
191 The sample below demonstrates how to use RotatedRect:
195 Mat image(200, 200, CV_8UC3, Scalar(0));
196 RotatedRect rRect = RotatedRect(Point2f(100,100), Size2f(100,50), 30);
199 rRect.points(vertices);
200 for (int i = 0; i < 4; i++)
201 line(image, vertices[i], vertices[(i+1)%4], Scalar(0,255,0));
203 Rect brect = rRect.boundingRect();
204 rectangle(image, brect, Scalar(255,0,0));
206 imshow("rectangles", image);
209 .. image:: pics/rotatedrect.png
213 :ocv:func:`CamShift` ,
214 :ocv:func:`fitEllipse` ,
215 :ocv:func:`minAreaRect` ,
216 :ocv:struct:`CvBox2D`
220 .. ocv:class:: TermCriteria
222 The class defining termination criteria for iterative algorithms. You can initialize it by default constructor and then override any parameters, or the structure may be fully initialized using the advanced variant of the constructor.
224 TermCriteria::TermCriteria
225 --------------------------
228 .. ocv:function:: TermCriteria::TermCriteria()
230 .. ocv:function:: TermCriteria::TermCriteria(int type, int maxCount, double epsilon)
232 .. ocv:function:: TermCriteria::TermCriteria(const CvTermCriteria& criteria)
234 :param type: The type of termination criteria: ``TermCriteria::COUNT``, ``TermCriteria::EPS`` or ``TermCriteria::COUNT`` + ``TermCriteria::EPS``.
236 :param maxCount: The maximum number of iterations or elements to compute.
238 :param epsilon: The desired accuracy or change in parameters at which the iterative algorithm stops.
240 :param criteria: Termination criteria in the deprecated ``CvTermCriteria`` format.
242 TermCriteria::operator CvTermCriteria
243 -------------------------------------
244 Converts to the deprecated ``CvTermCriteria`` format.
246 .. ocv:function:: TermCriteria::operator CvTermCriteria() const
252 Template class for small matrices whose type and size are known at compilation time: ::
254 template<typename _Tp, int m, int n> class Matx {...};
256 typedef Matx<float, 1, 2> Matx12f;
257 typedef Matx<double, 1, 2> Matx12d;
259 typedef Matx<float, 1, 6> Matx16f;
260 typedef Matx<double, 1, 6> Matx16d;
262 typedef Matx<float, 2, 1> Matx21f;
263 typedef Matx<double, 2, 1> Matx21d;
265 typedef Matx<float, 6, 1> Matx61f;
266 typedef Matx<double, 6, 1> Matx61d;
268 typedef Matx<float, 2, 2> Matx22f;
269 typedef Matx<double, 2, 2> Matx22d;
271 typedef Matx<float, 6, 6> Matx66f;
272 typedef Matx<double, 6, 6> Matx66d;
274 If you need a more flexible type, use :ocv:class:`Mat` . The elements of the matrix ``M`` are accessible using the ``M(i,j)`` notation. Most of the common matrix operations (see also
275 :ref:`MatrixExpressions` ) are available. To do an operation on ``Matx`` that is not implemented, you can easily convert the matrix to
276 ``Mat`` and backwards. ::
281 cout << sum(Mat(m*m.t())) << endl;
288 Template class for short numerical vectors, a partial case of :ocv:class:`Matx`: ::
290 template<typename _Tp, int n> class Vec : public Matx<_Tp, n, 1> {...};
292 typedef Vec<uchar, 2> Vec2b;
293 typedef Vec<uchar, 3> Vec3b;
294 typedef Vec<uchar, 4> Vec4b;
296 typedef Vec<short, 2> Vec2s;
297 typedef Vec<short, 3> Vec3s;
298 typedef Vec<short, 4> Vec4s;
300 typedef Vec<int, 2> Vec2i;
301 typedef Vec<int, 3> Vec3i;
302 typedef Vec<int, 4> Vec4i;
304 typedef Vec<float, 2> Vec2f;
305 typedef Vec<float, 3> Vec3f;
306 typedef Vec<float, 4> Vec4f;
307 typedef Vec<float, 6> Vec6f;
309 typedef Vec<double, 2> Vec2d;
310 typedef Vec<double, 3> Vec3d;
311 typedef Vec<double, 4> Vec4d;
312 typedef Vec<double, 6> Vec6d;
314 It is possible to convert ``Vec<T,2>`` to/from ``Point_``, ``Vec<T,3>`` to/from ``Point3_`` , and ``Vec<T,4>`` to :ocv:struct:`CvScalar` or :ocv:class:`Scalar_`. Use ``operator[]`` to access the elements of ``Vec``.
316 All the expected vector operations are also implemented:
320 * ``v1 = v2 * scale``
321 * ``v1 = scale * v2``
323 * ``v1 += v2`` and other augmenting operations
324 * ``v1 == v2, v1 != v2``
325 * ``norm(v1)`` (euclidean norm)
327 The ``Vec`` class is commonly used to describe pixel types of multi-channel arrays. See :ocv:class:`Mat` for details.
331 .. ocv:class:: Scalar_
333 Template class for a 4-element vector derived from Vec. ::
335 template<typename _Tp> class Scalar_ : public Vec<_Tp, 4> { ... };
337 typedef Scalar_<double> Scalar;
339 Being derived from ``Vec<_Tp, 4>`` , ``Scalar_`` and ``Scalar`` can be used just as typical 4-element vectors. In addition, they can be converted to/from ``CvScalar`` . The type ``Scalar`` is widely used in OpenCV to pass pixel values.
345 Template class specifying a continuous subsequence (slice) of a sequence. ::
354 The class is used to specify a row or a column span in a matrix (
355 :ocv:class:`Mat` ) and for many other purposes. ``Range(a,b)`` is basically the same as ``a:b`` in Matlab or ``a..b`` in Python. As in Python, ``start`` is an inclusive left boundary of the range and ``end`` is an exclusive right boundary of the range. Such a half-opened interval is usually denoted as
356 :math:`[start,end)` .
358 The static method ``Range::all()`` returns a special variable that means "the whole sequence" or "the whole range", just like " ``:`` " in Matlab or " ``...`` " in Python. All the methods and functions in OpenCV that take ``Range`` support this special ``Range::all()`` value. But, of course, in case of your own custom processing, you will probably have to check and handle it explicitly: ::
360 void my_function(..., const Range& r, ....)
362 if(r == Range::all()) {
363 // process all the data
366 // process [r.start, r.end)
377 Template class for smart reference-counting pointers ::
379 template<typename _Tp> class Ptr
382 // default constructor
384 // constructor that wraps the object pointer
386 // destructor: calls release()
388 // copy constructor; increments ptr's reference counter
390 // assignment operator; decrements own reference counter
391 // (with release()) and increments ptr's reference counter
392 Ptr& operator = (const Ptr& ptr);
393 // increments reference counter
395 // decrements reference counter; when it becomes 0,
396 // delete_obj() is called
398 // user-specified custom object deletion operation.
399 // by default, "delete obj;" is called
401 // returns true if obj == 0;
404 // provide access to the object fields and methods
406 const _Tp* operator -> () const;
408 // return the underlying object pointer;
409 // thanks to the methods, the Ptr<_Tp> can be
410 // used instead of _Tp*
412 operator const _Tp*() const;
414 // the encapsulated object pointer
416 // the associated reference counter
421 The ``Ptr<_Tp>`` class is a template class that wraps pointers of the corresponding type. It is
422 similar to ``shared_ptr`` that is part of the Boost library
423 (http://www.boost.org/doc/libs/1_40_0/libs/smart_ptr/shared_ptr.htm) and also part of the
424 `C++0x <http://en.wikipedia.org/wiki/C++0x>`_ standard.
426 This class provides the following options:
429 Default constructor, copy constructor, and assignment operator for an arbitrary C++ class
430 or a C structure. For some objects, like files, windows, mutexes, sockets, and others, a copy
431 constructor or an assignment operator are difficult to define. For some other objects, like
432 complex classifiers in OpenCV, copy constructors are absent and not easy to implement. Finally,
433 some of complex OpenCV and your own data structures may be written in C.
434 However, copy constructors and default constructors can simplify programming a lot.Besides,
435 they are often required (for example, by STL containers). By wrapping a pointer to such a
436 complex object ``TObj`` to ``Ptr<TObj>``, you automatically get all of the necessary
437 constructors and the assignment operator.
440 *O(1)* complexity of the above-mentioned operations. While some structures, like ``std::vector``,
441 provide a copy constructor and an assignment operator, the operations may take a considerable
442 amount of time if the data structures are large. But if the structures are put into ``Ptr<>``,
443 the overhead is small and independent of the data size.
446 Automatic destruction, even for C structures. See the example below with ``FILE*``.
449 Heterogeneous collections of objects. The standard STL and most other C++ and OpenCV containers
450 can store only objects of the same type and the same size. The classical solution to store objects
451 of different types in the same container is to store pointers to the base class ``base_class_t*``
452 instead but then you loose the automatic memory management. Again, by using ``Ptr<base_class_t>()``
453 instead of the raw pointers, you can solve the problem.
455 The ``Ptr`` class treats the wrapped object as a black box. The reference counter is allocated and
456 managed separately. The only thing the pointer class needs to know about the object is how to
457 deallocate it. This knowledge is encapsulated in the ``Ptr::delete_obj()`` method that is called when
458 the reference counter becomes 0. If the object is a C++ class instance, no additional coding is
459 needed, because the default implementation of this method calls ``delete obj;``. However, if the
460 object is deallocated in a different way, the specialized method should be created. For example,
461 if you want to wrap ``FILE``, the ``delete_obj`` may be implemented as follows: ::
463 template<> inline void Ptr<FILE>::delete_obj()
465 fclose(obj); // no need to clear the pointer afterwards,
466 // it is done externally.
471 Ptr<FILE> f(fopen("myfile.txt", "r"));
476 // the file will be closed automatically by the Ptr<FILE> destructor.
479 .. note:: The reference increment/decrement operations are implemented as atomic operations,
480 and therefore it is normally safe to use the classes in multi-threaded applications.
481 The same is true for :ocv:class:`Mat` and other C++ OpenCV classes that operate on
482 the reference counters.
486 Various Ptr constructors.
488 .. ocv:function:: Ptr::Ptr()
489 .. ocv:function:: Ptr::Ptr(_Tp* _obj)
490 .. ocv:function:: Ptr::Ptr(const Ptr& ptr)
492 :param _obj: Object for copy.
493 :param ptr: Object for copy.
499 .. ocv:function:: Ptr::~Ptr()
505 .. ocv:function:: Ptr& Ptr::operator = (const Ptr& ptr)
507 :param ptr: Object for assignment.
509 Decrements own reference counter (with ``release()``) and increments ptr's reference counter.
513 Increments reference counter.
515 .. ocv:function:: void Ptr::addref()
519 Decrements reference counter; when it becomes 0, ``delete_obj()`` is called.
521 .. ocv:function:: void Ptr::release()
525 User-specified custom object deletion operation. By default, ``delete obj;`` is called.
527 .. ocv:function:: void Ptr::delete_obj()
531 Returns true if obj == 0;
537 Provide access to the object fields and methods.
539 .. ocv:function:: template<typename _Tp> _Tp* Ptr::operator -> ()
540 .. ocv:function:: template<typename _Tp> const _Tp* Ptr::operator -> () const
545 Returns the underlying object pointer. Thanks to the methods, the ``Ptr<_Tp>`` can be used instead
548 .. ocv:function:: template<typename _Tp> Ptr::operator _Tp* ()
549 .. ocv:function:: template<typename _Tp> Ptr::operator const _Tp*() const
556 OpenCV C++ n-dimensional dense array class ::
561 // ... a lot of methods ...
564 /*! includes several bit-fields:
565 - the magic signature
571 //! the array dimensionality, >= 2
573 //! the number of rows and columns or (-1, -1) when the array has more than 2 dimensions
575 //! pointer to the data
578 //! pointer to the reference counter;
579 // when array points to user-allocated data, the pointer is NULL
587 The class ``Mat`` represents an n-dimensional dense numerical single-channel or multi-channel array. It can be used to store real or complex-valued vectors and matrices, grayscale or color images, voxel volumes, vector fields, point clouds, tensors, histograms (though, very high-dimensional histograms may be better stored in a ``SparseMat`` ). The data layout of the array
588 :math:`M` is defined by the array ``M.step[]``, so that the address of element
589 :math:`(i_0,...,i_{M.dims-1})`, where
590 :math:`0\leq i_k<M.size[k]`, is computed as:
594 addr(M_{i_0,...,i_{M.dims-1}}) = M.data + M.step[0]*i_0 + M.step[1]*i_1 + ... + M.step[M.dims-1]*i_{M.dims-1}
596 In case of a 2-dimensional array, the above formula is reduced to:
600 addr(M_{i,j}) = M.data + M.step[0]*i + M.step[1]*j
602 Note that ``M.step[i] >= M.step[i+1]`` (in fact, ``M.step[i] >= M.step[i+1]*M.size[i+1]`` ). This means that 2-dimensional matrices are stored row-by-row, 3-dimensional matrices are stored plane-by-plane, and so on. ``M.step[M.dims-1]`` is minimal and always equal to the element size ``M.elemSize()`` .
604 So, the data layout in ``Mat`` is fully compatible with ``CvMat``, ``IplImage``, and ``CvMatND`` types from OpenCV 1.x. It is also compatible with the majority of dense array types from the standard toolkits and SDKs, such as Numpy (ndarray), Win32 (independent device bitmaps), and others, that is, with any array that uses *steps* (or *strides*) to compute the position of a pixel. Due to this compatibility, it is possible to make a ``Mat`` header for user-allocated data and process it in-place using OpenCV functions.
606 There are many different ways to create a ``Mat`` object. The most popular options are listed below:
610 Use the ``create(nrows, ncols, type)`` method or the similar ``Mat(nrows, ncols, type[, fillValue])`` constructor. A new array of the specified size and type is allocated. ``type`` has the same meaning as in the ``cvCreateMat`` method.
611 For example, ``CV_8UC1`` means a 8-bit single-channel array, ``CV_32FC2`` means a 2-channel (complex) floating-point array, and so on.
615 // make a 7x7 complex matrix filled with 1+3j.
616 Mat M(7,7,CV_32FC2,Scalar(1,3));
617 // and now turn M to a 100x60 15-channel 8-bit matrix.
618 // The old content will be deallocated
619 M.create(100,60,CV_8UC(15));
623 As noted in the introduction to this chapter, ``create()`` allocates only a new array when the shape or type of the current array are different from the specified ones.
627 Create a multi-dimensional array:
631 // create a 100x100x100 8-bit array
632 int sz[] = {100, 100, 100};
633 Mat bigCube(3, sz, CV_8U, Scalar::all(0));
637 It passes the number of dimensions =1 to the ``Mat`` constructor but the created array will be 2-dimensional with the number of columns set to 1. So, ``Mat::dims`` is always >= 2 (can also be 0 when the array is empty).
641 Use a copy constructor or assignment operator where there can be an array or expression on the right side (see below). As noted in the introduction, the array assignment is an O(1) operation because it only copies the header and increases the reference counter. The ``Mat::clone()`` method can be used to get a full (deep) copy of the array when you need it.
645 Construct a header for a part of another array. It can be a single row, single column, several rows, several columns, rectangular region in the array (called a *minor* in algebra) or a diagonal. Such operations are also O(1) because the new header references the same data. You can actually modify a part of the array using this feature, for example:
649 // add the 5-th row, multiplied by 3 to the 3rd row
650 M.row(3) = M.row(3) + M.row(5)*3;
652 // now copy the 7-th column to the 1-st column
653 // M.col(1) = M.col(7); // this will not work
657 // create a new 320x240 image
658 Mat img(Size(320,240),CV_8UC3);
660 Mat roi(img, Rect(10,10,100,100));
661 // fill the ROI with (0,255,0) (which is green in RGB space);
662 // the original 320x240 image will be modified
663 roi = Scalar(0,255,0);
667 Due to the additional ``datastart`` and ``dataend`` members, it is possible to compute a relative sub-array position in the main *container* array using ``locateROI()``:
671 Mat A = Mat::eye(10, 10, CV_32S);
672 // extracts A columns, 1 (inclusive) to 3 (exclusive).
673 Mat B = A(Range::all(), Range(1, 3));
674 // extracts B rows, 5 (inclusive) to 9 (exclusive).
675 // that is, C ~ A(Range(5, 9), Range(1, 3))
676 Mat C = B(Range(5, 9), Range::all());
677 Size size; Point ofs;
678 C.locateROI(size, ofs);
679 // size will be (width=10,height=10) and the ofs will be (x=1, y=5)
683 As in case of whole matrices, if you need a deep copy, use the ``clone()`` method of the extracted sub-matrices.
687 Make a header for user-allocated data. It can be useful to do the following:
690 Process "foreign" data using OpenCV (for example, when you implement a DirectShow* filter or a processing module for ``gstreamer``, and so on). For example:
694 void process_video_frame(const unsigned char* pixels,
695 int width, int height, int step)
697 Mat img(height, width, CV_8UC3, pixels, step);
698 GaussianBlur(img, img, Size(7,7), 1.5, 1.5);
704 Quickly initialize small matrices and/or get a super-fast element access.
708 double m[3][3] = {{a, b, c}, {d, e, f}, {g, h, i}};
709 Mat M = Mat(3, 3, CV_64F, m).inv();
713 Partial yet very common cases of this *user-allocated data* case are conversions from ``CvMat`` and ``IplImage`` to ``Mat``. For this purpose, there are special constructors taking pointers to ``CvMat`` or ``IplImage`` and the optional flag indicating whether to copy the data or not.
715 Backward conversion from ``Mat`` to ``CvMat`` or ``IplImage`` is provided via cast operators ``Mat::operator CvMat() const`` and ``Mat::operator IplImage()``. The operators do NOT copy the data.
719 IplImage* img = cvLoadImage("greatwave.jpg", 1);
720 Mat mtx(img); // convert IplImage* -> Mat
721 CvMat oldmat = mtx; // convert Mat -> CvMat
722 CV_Assert(oldmat.cols == img->width && oldmat.rows == img->height &&
723 oldmat.data.ptr == (uchar*)img->imageData && oldmat.step == img->widthStep);
729 Use MATLAB-style array initializers, ``zeros(), ones(), eye()``, for example:
733 // create a double-precision identity martix and add it to M.
734 M += Mat::eye(M.rows, M.cols, CV_64F);
740 Use a comma-separated initializer:
744 // create a 3x3 double-precision identity matrix
745 Mat M = (Mat_<double>(3,3) << 1, 0, 0, 0, 1, 0, 0, 0, 1);
749 With this approach, you first call a constructor of the :ocv:class:`Mat_` class with the proper parameters, and then you just put ``<<`` operator followed by comma-separated values that can be constants, variables, expressions, and so on. Also, note the extra parentheses required to avoid compilation errors.
751 Once the array is created, it is automatically managed via a reference-counting mechanism. If the array header is built on top of user-allocated data, you should handle the data by yourself.
752 The array data is deallocated when no one points to it. If you want to release the data pointed by a array header before the array destructor is called, use ``Mat::release()`` .
754 The next important thing to learn about the array class is element access. This manual already described how to compute an address of each array element. Normally, you are not required to use the formula directly in the code. If you know the array element type (which can be retrieved using the method ``Mat::type()`` ), you can access the element
755 :math:`M_{ij}` of a 2-dimensional array as: ::
757 M.at<double>(i,j) += 1.f;
760 assuming that M is a double-precision floating-point array. There are several variants of the method ``at`` for a different number of dimensions.
762 If you need to process a whole row of a 2D array, the most efficient way is to get the pointer to the row first, and then just use the plain C operator ``[]`` : ::
764 // compute sum of positive matrix elements
765 // (assuming that M isa double-precision matrix)
767 for(int i = 0; i < M.rows; i++)
769 const double* Mi = M.ptr<double>(i);
770 for(int j = 0; j < M.cols; j++)
771 sum += std::max(Mi[j], 0.);
775 Some operations, like the one above, do not actually depend on the array shape. They just process elements of an array one by one (or elements from multiple arrays that have the same coordinates, for example, array addition). Such operations are called *element-wise*. It makes sense to check whether all the input/output arrays are continuous, namely, have no gaps at the end of each row. If yes, process them as a long single row: ::
777 // compute the sum of positive matrix elements, optimized variant
779 int cols = M.cols, rows = M.rows;
785 for(int i = 0; i < rows; i++)
787 const double* Mi = M.ptr<double>(i);
788 for(int j = 0; j < cols; j++)
789 sum += std::max(Mi[j], 0.);
793 In case of the continuous matrix, the outer loop body is executed just once. So, the overhead is smaller, which is especially noticeable in case of small matrices.
795 Finally, there are STL-style iterators that are smart enough to skip gaps between successive rows: ::
797 // compute sum of positive matrix elements, iterator-based variant
799 MatConstIterator_<double> it = M.begin<double>(), it_end = M.end<double>();
800 for(; it != it_end; ++it)
801 sum += std::max(*it, 0.);
804 The matrix iterators are random-access iterators, so they can be passed to any STL algorithm, including ``std::sort()`` .
808 * An example demonstrating the serial out capabilities of cv::Mat can be found at opencv_source_code/samples/cpp/cout_mat.cpp
810 .. _MatrixExpressions:
815 This is a list of implemented matrix operations that can be combined in arbitrary complex expressions
816 (here ``A``, ``B`` stand for matrices ( ``Mat`` ), ``s`` for a scalar ( ``Scalar`` ),
817 ``alpha`` for a real-valued scalar ( ``double`` )):
820 Addition, subtraction, negation:
821 ``A+B, A-B, A+s, A-s, s+A, s-A, -A``
828 Per-element multiplication and division:
829 ``A.mul(B), A/B, alpha/A``
832 Matrix multiplication:
837 ``A.t()`` (means ``A``\ :sup:`T`)
840 Matrix inversion and pseudo-inversion, solving linear systems and least-squares problems:
842 ``A.inv([method])`` (~ ``A``\ :sup:`-1`) ``, A.inv([method])*B`` (~ ``X: AX=B``)
846 ``A cmpop B, A cmpop alpha, alpha cmpop A``, where ``cmpop`` is one of ``: >, >=, ==, !=, <=, <``. The result of comparison is an 8-bit single channel mask whose elements are set to 255 (if the particular element or pair of elements satisfy the condition) or 0.
849 Bitwise logical operations: ``A logicop B, A logicop s, s logicop A, ~A``, where ``logicop`` is one of ``: &, |, ^``.
852 Element-wise minimum and maximum:
853 ``min(A, B), min(A, alpha), max(A, B), max(A, alpha)``
856 Element-wise absolute value:
860 Cross-product, dot-product:
865 Any function of matrix or matrices and scalars that returns a matrix or a scalar, such as ``norm``, ``mean``, ``sum``, ``countNonZero``, ``trace``, ``determinant``, ``repeat``, and others.
868 Matrix initializers ( ``Mat::eye(), Mat::zeros(), Mat::ones()`` ), matrix comma-separated initializers, matrix constructors and operators that extract sub-matrices (see :ocv:class:`Mat` description).
871 ``Mat_<destination_type>()`` constructors to cast the result to the proper type.
873 .. note:: Comma-separated initializers and probably some other operations may require additional explicit ``Mat()`` or ``Mat_<T>()`` constructor calls to resolve a possible ambiguity.
875 Here are examples of matrix expressions:
879 // compute pseudo-inverse of A, equivalent to A.inv(DECOMP_SVD)
881 Mat pinvA = svd.vt.t()*Mat::diag(1./svd.w)*svd.u.t();
883 // compute the new vector of parameters in the Levenberg-Marquardt algorithm
884 x -= (A.t()*A + lambda*Mat::eye(A.cols,A.cols,A.type())).inv(DECOMP_CHOLESKY)*(A.t()*err);
886 // sharpen image using "unsharp mask" algorithm
887 Mat blurred; double sigma = 1, threshold = 5, amount = 1;
888 GaussianBlur(img, blurred, Size(), sigma, sigma);
889 Mat lowConstrastMask = abs(img - blurred) < threshold;
890 Mat sharpened = img*(1+amount) + blurred*(-amount);
891 img.copyTo(sharpened, lowContrastMask);
896 Below is the formal description of the ``Mat`` methods.
900 Various Mat constructors
902 .. ocv:function:: Mat::Mat()
904 .. ocv:function:: Mat::Mat(int rows, int cols, int type)
906 .. ocv:function:: Mat::Mat(Size size, int type)
908 .. ocv:function:: Mat::Mat(int rows, int cols, int type, const Scalar& s)
910 .. ocv:function:: Mat::Mat(Size size, int type, const Scalar& s)
912 .. ocv:function:: Mat::Mat(const Mat& m)
914 .. ocv:function:: Mat::Mat(int rows, int cols, int type, void* data, size_t step=AUTO_STEP)
916 .. ocv:function:: Mat::Mat(Size size, int type, void* data, size_t step=AUTO_STEP)
918 .. ocv:function:: Mat::Mat( const Mat& m, const Range& rowRange, const Range& colRange=Range::all() )
920 .. ocv:function:: Mat::Mat(const Mat& m, const Rect& roi)
922 .. ocv:function:: Mat::Mat(const CvMat* m, bool copyData=false)
924 .. ocv:function:: Mat::Mat(const IplImage* img, bool copyData=false)
926 .. ocv:function:: template<typename T, int n> explicit Mat::Mat(const Vec<T, n>& vec, bool copyData=true)
928 .. ocv:function:: template<typename T, int m, int n> explicit Mat::Mat(const Matx<T, m, n>& vec, bool copyData=true)
930 .. ocv:function:: template<typename T> explicit Mat::Mat(const vector<T>& vec, bool copyData=false)
932 .. ocv:function:: Mat::Mat(int ndims, const int* sizes, int type)
934 .. ocv:function:: Mat::Mat(int ndims, const int* sizes, int type, const Scalar& s)
936 .. ocv:function:: Mat::Mat(int ndims, const int* sizes, int type, void* data, const size_t* steps=0)
938 .. ocv:function:: Mat::Mat(const Mat& m, const Range* ranges)
940 :param ndims: Array dimensionality.
942 :param rows: Number of rows in a 2D array.
944 :param cols: Number of columns in a 2D array.
946 :param roi: Region of interest.
948 :param size: 2D array size: ``Size(cols, rows)`` . In the ``Size()`` constructor, the number of rows and the number of columns go in the reverse order.
950 :param sizes: Array of integers specifying an n-dimensional array shape.
952 :param type: Array type. Use ``CV_8UC1, ..., CV_64FC4`` to create 1-4 channel matrices, or ``CV_8UC(n), ..., CV_64FC(n)`` to create multi-channel (up to ``CV_MAX_CN`` channels) matrices.
954 :param s: An optional value to initialize each matrix element with. To set all the matrix elements to the particular value after the construction, use the assignment operator ``Mat::operator=(const Scalar& value)`` .
956 :param data: Pointer to the user data. Matrix constructors that take ``data`` and ``step`` parameters do not allocate matrix data. Instead, they just initialize the matrix header that points to the specified data, which means that no data is copied. This operation is very efficient and can be used to process external data using OpenCV functions. The external data is not automatically deallocated, so you should take care of it.
958 :param step: Number of bytes each matrix row occupies. The value should include the padding bytes at the end of each row, if any. If the parameter is missing (set to ``AUTO_STEP`` ), no padding is assumed and the actual step is calculated as ``cols*elemSize()`` . See :ocv:func:`Mat::elemSize` .
960 :param steps: Array of ``ndims-1`` steps in case of a multi-dimensional array (the last step is always set to the element size). If not specified, the matrix is assumed to be continuous.
962 :param m: Array that (as a whole or partly) is assigned to the constructed matrix. No data is copied by these constructors. Instead, the header pointing to ``m`` data or its sub-array is constructed and associated with it. The reference counter, if any, is incremented. So, when you modify the matrix formed using such a constructor, you also modify the corresponding elements of ``m`` . If you want to have an independent copy of the sub-array, use ``Mat::clone()`` .
964 :param img: Pointer to the old-style ``IplImage`` image structure. By default, the data is shared between the original image and the new matrix. But when ``copyData`` is set, the full copy of the image data is created.
966 :param vec: STL vector whose elements form the matrix. The matrix has a single column and the number of rows equal to the number of vector elements. Type of the matrix matches the type of vector elements. The constructor can handle arbitrary types, for which there is a properly declared :ocv:class:`DataType` . This means that the vector elements must be primitive numbers or uni-type numerical tuples of numbers. Mixed-type structures are not supported. The corresponding constructor is explicit. Since STL vectors are not automatically converted to ``Mat`` instances, you should write ``Mat(vec)`` explicitly. Unless you copy the data into the matrix ( ``copyData=true`` ), no new elements will be added to the vector because it can potentially yield vector data reallocation, and, thus, the matrix data pointer will be invalid.
968 :param copyData: Flag to specify whether the underlying data of the STL vector or the old-style ``CvMat`` or ``IplImage`` should be copied to (``true``) or shared with (``false``) the newly constructed matrix. When the data is copied, the allocated buffer is managed using ``Mat`` reference counting mechanism. While the data is shared, the reference counter is NULL, and you should not deallocate the data until the matrix is not destructed.
970 :param rowRange: Range of the ``m`` rows to take. As usual, the range start is inclusive and the range end is exclusive. Use ``Range::all()`` to take all the rows.
972 :param colRange: Range of the ``m`` columns to take. Use ``Range::all()`` to take all the columns.
974 :param ranges: Array of selected ranges of ``m`` along each dimensionality.
976 These are various constructors that form a matrix. As noted in the :ref:`AutomaticAllocation`,
977 often the default constructor is enough, and the proper matrix will be allocated by an OpenCV function. The constructed matrix can further be assigned to another matrix or matrix expression or can be allocated with
978 :ocv:func:`Mat::create` . In the former case, the old content is de-referenced.
985 .. ocv:function:: Mat::~Mat()
987 The matrix destructor calls :ocv:func:`Mat::release` .
992 Provides matrix assignment operators.
994 .. ocv:function:: Mat& Mat::operator = (const Mat& m)
996 .. ocv:function:: Mat& Mat::operator =( const MatExpr& expr )
998 .. ocv:function:: Mat& Mat::operator = (const Scalar& s)
1000 :param m: Assigned, right-hand-side matrix. Matrix assignment is an O(1) operation. This means that no data is copied but the data is shared and the reference counter, if any, is incremented. Before assigning new data, the old data is de-referenced via :ocv:func:`Mat::release` .
1002 :param expr: Assigned matrix expression object. As opposite to the first form of the assignment operation, the second form can reuse already allocated matrix if it has the right size and type to fit the matrix expression result. It is automatically handled by the real function that the matrix expressions is expanded to. For example, ``C=A+B`` is expanded to ``add(A, B, C)``, and :func:`add` takes care of automatic ``C`` reallocation.
1004 :param s: Scalar assigned to each matrix element. The matrix size or type is not changed.
1006 These are available assignment operators. Since they all are very different, make sure to read the operator parameters description.
1010 Creates a matrix header for the specified matrix row.
1012 .. ocv:function:: Mat Mat::row(int y) const
1014 :param y: A 0-based row index.
1016 The method makes a new header for the specified matrix row and returns it. This is an O(1) operation, regardless of the matrix size. The underlying data of the new matrix is shared with the original matrix. Here is the example of one of the classical basic matrix processing operations, ``axpy``, used by LU and many other algorithms: ::
1018 inline void matrix_axpy(Mat& A, int i, int j, double alpha)
1020 A.row(i) += A.row(j)*alpha;
1026 In the current implementation, the following code does not work as expected: ::
1030 A.row(i) = A.row(j); // will not work
1033 This happens because ``A.row(i)`` forms a temporary header that is further assigned to another header. Remember that each of these operations is O(1), that is, no data is copied. Thus, the above assignment is not true if you may have expected the j-th row to be copied to the i-th row. To achieve that, you should either turn this simple assignment into an expression or use the :ocv:func:`Mat::copyTo` method: ::
1037 // works, but looks a bit obscure.
1038 A.row(i) = A.row(j) + 0;
1040 // this is a bit longer, but the recommended method.
1041 A.row(j).copyTo(A.row(i));
1045 Creates a matrix header for the specified matrix column.
1047 .. ocv:function:: Mat Mat::col(int x) const
1049 :param x: A 0-based column index.
1051 The method makes a new header for the specified matrix column and returns it. This is an O(1) operation, regardless of the matrix size. The underlying data of the new matrix is shared with the original matrix. See also the
1052 :ocv:func:`Mat::row` description.
1057 Creates a matrix header for the specified row span.
1059 .. ocv:function:: Mat Mat::rowRange(int startrow, int endrow) const
1061 .. ocv:function:: Mat Mat::rowRange(const Range& r) const
1063 :param startrow: An inclusive 0-based start index of the row span.
1065 :param endrow: An exclusive 0-based ending index of the row span.
1067 :param r: :ocv:class:`Range` structure containing both the start and the end indices.
1069 The method makes a new header for the specified row span of the matrix. Similarly to
1070 :ocv:func:`Mat::row` and
1071 :ocv:func:`Mat::col` , this is an O(1) operation.
1075 Creates a matrix header for the specified row span.
1077 .. ocv:function:: Mat Mat::colRange(int startcol, int endcol) const
1079 .. ocv:function:: Mat Mat::colRange(const Range& r) const
1081 :param startcol: An inclusive 0-based start index of the column span.
1083 :param endcol: An exclusive 0-based ending index of the column span.
1085 :param r: :ocv:class:`Range` structure containing both the start and the end indices.
1087 The method makes a new header for the specified column span of the matrix. Similarly to
1088 :ocv:func:`Mat::row` and
1089 :ocv:func:`Mat::col` , this is an O(1) operation.
1093 Extracts a diagonal from a matrix, or creates a diagonal matrix.
1095 .. ocv:function:: Mat Mat::diag( int d=0 ) const
1097 .. ocv:function:: static Mat Mat::diag( const Mat& d )
1099 :param d: Single-column matrix that forms a diagonal matrix or index of the diagonal, with the following values:
1101 * **d=0** is the main diagonal.
1103 * **d>0** is a diagonal from the lower half. For example, ``d=1`` means the diagonal is set immediately below the main one.
1105 * **d<0** is a diagonal from the upper half. For example, ``d=1`` means the diagonal is set immediately above the main one.
1107 The method makes a new header for the specified matrix diagonal. The new matrix is represented as a single-column matrix. Similarly to
1108 :ocv:func:`Mat::row` and
1109 :ocv:func:`Mat::col` , this is an O(1) operation.
1113 Creates a full copy of the array and the underlying data.
1115 .. ocv:function:: Mat Mat::clone() const
1117 The method creates a full copy of the array. The original ``step[]`` is not taken into account. So, the array copy is a continuous array occupying ``total()*elemSize()`` bytes.
1122 Copies the matrix to another one.
1124 .. ocv:function:: void Mat::copyTo( OutputArray m ) const
1125 .. ocv:function:: void Mat::copyTo( OutputArray m, InputArray mask ) const
1127 :param m: Destination matrix. If it does not have a proper size or type before the operation, it is reallocated.
1129 :param mask: Operation mask. Its non-zero elements indicate which matrix elements need to be copied.
1131 The method copies the matrix data to another matrix. Before copying the data, the method invokes ::
1133 m.create(this->size(), this->type);
1136 so that the destination matrix is reallocated if needed. While ``m.copyTo(m);`` works flawlessly, the function does not handle the case of a partial overlap between the source and the destination matrices.
1138 When the operation mask is specified, and the ``Mat::create`` call shown above reallocated the matrix, the newly allocated matrix is initialized with all zeros before copying the data.
1144 Converts an array to another data type with optional scaling.
1146 .. ocv:function:: void Mat::convertTo( OutputArray m, int rtype, double alpha=1, double beta=0 ) const
1148 :param m: output matrix; if it does not have a proper size or type before the operation, it is reallocated.
1150 :param rtype: desired output matrix type or, rather, the depth since the number of channels are the same as the input has; if ``rtype`` is negative, the output matrix will have the same type as the input.
1152 :param alpha: optional scale factor.
1154 :param beta: optional delta added to the scaled values.
1156 The method converts source pixel values to the target data type. ``saturate_cast<>`` is applied at the end to avoid possible overflows:
1160 m(x,y) = saturate \_ cast<rType>( \alpha (*this)(x,y) + \beta )
1165 Provides a functional form of ``convertTo``.
1167 .. ocv:function:: void Mat::assignTo( Mat& m, int type=-1 ) const
1169 :param m: Destination array.
1171 :param type: Desired destination array depth (or -1 if it should be the same as the source type).
1173 This is an internally used method called by the
1174 :ref:`MatrixExpressions` engine.
1178 Sets all or some of the array elements to the specified value.
1180 .. ocv:function:: Mat& Mat::setTo( InputArray value, InputArray mask=noArray() )
1182 :param value: Assigned scalar converted to the actual array type.
1184 :param mask: Operation mask of the same size as ``*this``. This is an advanced variant of the ``Mat::operator=(const Scalar& s)`` operator.
1189 Changes the shape and/or the number of channels of a 2D matrix without copying the data.
1191 .. ocv:function:: Mat Mat::reshape(int cn, int rows=0) const
1193 :param cn: New number of channels. If the parameter is 0, the number of channels remains the same.
1195 :param rows: New number of rows. If the parameter is 0, the number of rows remains the same.
1197 The method makes a new matrix header for ``*this`` elements. The new matrix may have a different size and/or different number of channels. Any combination is possible if:
1200 No extra elements are included into the new matrix and no elements are excluded. Consequently, the product ``rows*cols*channels()`` must stay the same after the transformation.
1203 No data is copied. That is, this is an O(1) operation. Consequently, if you change the number of rows, or the operation changes the indices of elements row in some other way, the matrix must be continuous. See
1204 :ocv:func:`Mat::isContinuous` .
1206 For example, if there is a set of 3D points stored as an STL vector, and you want to represent the points as a ``3xN`` matrix, do the following: ::
1208 std::vector<Point3f> vec;
1211 Mat pointMat = Mat(vec). // convert vector to Mat, O(1) operation
1212 reshape(1). // make Nx3 1-channel matrix out of Nx1 3-channel.
1213 // Also, an O(1) operation
1214 t(); // finally, transpose the Nx3 matrix.
1215 // This involves copying all the elements
1222 Transposes a matrix.
1224 .. ocv:function:: MatExpr Mat::t() const
1226 The method performs matrix transposition by means of matrix expressions. It does not perform the actual transposition but returns a temporary matrix transposition object that can be further used as a part of more complex matrix expressions or can be assigned to a matrix: ::
1228 Mat A1 = A + Mat::eye(A.size(), A.type)*lambda;
1229 Mat C = A1.t()*A1; // compute (A + lambda*I)^t * (A + lamda*I)
1236 .. ocv:function:: MatExpr Mat::inv(int method=DECOMP_LU) const
1238 :param method: Matrix inversion method. Possible values are the following:
1240 * **DECOMP_LU** is the LU decomposition. The matrix must be non-singular.
1242 * **DECOMP_CHOLESKY** is the Cholesky :math:`LL^T` decomposition for symmetrical positively defined matrices only. This type is about twice faster than LU on big matrices.
1244 * **DECOMP_SVD** is the SVD decomposition. If the matrix is singular or even non-square, the pseudo inversion is computed.
1246 The method performs a matrix inversion by means of matrix expressions. This means that a temporary matrix inversion object is returned by the method and can be used further as a part of more complex matrix expressions or can be assigned to a matrix.
1251 Performs an element-wise multiplication or division of the two matrices.
1253 .. ocv:function:: MatExpr Mat::mul(InputArray m, double scale=1) const
1255 :param m: Another array of the same type and the same size as ``*this``, or a matrix expression.
1257 :param scale: Optional scale factor.
1259 The method returns a temporary object encoding per-element array multiplication, with optional scale. Note that this is not a matrix multiplication that corresponds to a simpler "*" operator.
1263 Mat C = A.mul(5/B); // equivalent to divide(A, B, C, 5)
1268 Computes a cross-product of two 3-element vectors.
1270 .. ocv:function:: Mat Mat::cross(InputArray m) const
1272 :param m: Another cross-product operand.
1274 The method computes a cross-product of two 3-element vectors. The vectors must be 3-element floating-point vectors of the same shape and size. The result is another 3-element vector of the same shape and type as operands.
1279 Computes a dot-product of two vectors.
1281 .. ocv:function:: double Mat::dot(InputArray m) const
1283 :param m: another dot-product operand.
1285 The method computes a dot-product of two matrices. If the matrices are not single-column or single-row vectors, the top-to-bottom left-to-right scan ordering is used to treat them as 1D vectors. The vectors must have the same size and type. If the matrices have more than one channel, the dot products from all the channels are summed together.
1290 Returns a zero array of the specified size and type.
1292 .. ocv:function:: static MatExpr Mat::zeros(int rows, int cols, int type)
1293 .. ocv:function:: static MatExpr Mat::zeros(Size size, int type)
1294 .. ocv:function:: static MatExpr Mat::zeros( int ndims, const int* sz, int type )
1296 :param ndims: Array dimensionality.
1298 :param rows: Number of rows.
1300 :param cols: Number of columns.
1302 :param size: Alternative to the matrix size specification ``Size(cols, rows)`` .
1304 :param sz: Array of integers specifying the array shape.
1306 :param type: Created matrix type.
1308 The method returns a Matlab-style zero array initializer. It can be used to quickly form a constant array as a function parameter, part of a matrix expression, or as a matrix initializer. ::
1311 A = Mat::zeros(3, 3, CV_32F);
1314 In the example above, a new matrix is allocated only if ``A`` is not a 3x3 floating-point matrix. Otherwise, the existing matrix ``A`` is filled with zeros.
1319 Returns an array of all 1's of the specified size and type.
1321 .. ocv:function:: static MatExpr Mat::ones(int rows, int cols, int type)
1322 .. ocv:function:: static MatExpr Mat::ones(Size size, int type)
1323 .. ocv:function:: static MatExpr Mat::ones( int ndims, const int* sz, int type )
1325 :param ndims: Array dimensionality.
1327 :param rows: Number of rows.
1329 :param cols: Number of columns.
1331 :param size: Alternative to the matrix size specification ``Size(cols, rows)`` .
1333 :param sz: Array of integers specifying the array shape.
1335 :param type: Created matrix type.
1337 The method returns a Matlab-style 1's array initializer, similarly to
1338 :ocv:func:`Mat::zeros`. Note that using this method you can initialize an array with an arbitrary value, using the following Matlab idiom: ::
1340 Mat A = Mat::ones(100, 100, CV_8U)*3; // make 100x100 matrix filled with 3.
1343 The above operation does not form a 100x100 matrix of 1's and then multiply it by 3. Instead, it just remembers the scale factor (3 in this case) and use it when actually invoking the matrix initializer.
1348 Returns an identity matrix of the specified size and type.
1350 .. ocv:function:: static MatExpr Mat::eye(int rows, int cols, int type)
1351 .. ocv:function:: static MatExpr Mat::eye(Size size, int type)
1353 :param rows: Number of rows.
1355 :param cols: Number of columns.
1357 :param size: Alternative matrix size specification as ``Size(cols, rows)`` .
1359 :param type: Created matrix type.
1361 The method returns a Matlab-style identity matrix initializer, similarly to
1362 :ocv:func:`Mat::zeros`. Similarly to
1363 :ocv:func:`Mat::ones`, you can use a scale operation to create a scaled identity matrix efficiently: ::
1365 // make a 4x4 diagonal matrix with 0.1's on the diagonal.
1366 Mat A = Mat::eye(4, 4, CV_32F)*0.1;
1371 Allocates new array data if needed.
1373 .. ocv:function:: void Mat::create(int rows, int cols, int type)
1374 .. ocv:function:: void Mat::create(Size size, int type)
1375 .. ocv:function:: void Mat::create(int ndims, const int* sizes, int type)
1377 :param ndims: New array dimensionality.
1379 :param rows: New number of rows.
1381 :param cols: New number of columns.
1383 :param size: Alternative new matrix size specification: ``Size(cols, rows)``
1385 :param sizes: Array of integers specifying a new array shape.
1387 :param type: New matrix type.
1389 This is one of the key ``Mat`` methods. Most new-style OpenCV functions and methods that produce arrays call this method for each output array. The method uses the following algorithm:
1392 If the current array shape and the type match the new ones, return immediately. Otherwise, de-reference the previous data by calling
1393 :ocv:func:`Mat::release`.
1396 Initialize the new header.
1399 Allocate the new data of ``total()*elemSize()`` bytes.
1402 Allocate the new, associated with the data, reference counter and set it to 1.
1404 Such a scheme makes the memory management robust and efficient at the same time and helps avoid extra typing for you. This means that usually there is no need to explicitly allocate output arrays. That is, instead of writing: ::
1408 Mat gray(color.rows, color.cols, color.depth());
1409 cvtColor(color, gray, CV_BGR2GRAY);
1412 you can simply write: ::
1417 cvtColor(color, gray, CV_BGR2GRAY);
1420 because ``cvtColor`` , as well as the most of OpenCV functions, calls ``Mat::create()`` for the output array internally.
1425 Increments the reference counter.
1427 .. ocv:function:: void Mat::addref()
1429 The method increments the reference counter associated with the matrix data. If the matrix header points to an external data set (see
1430 :ocv:func:`Mat::Mat` ), the reference counter is NULL, and the method has no effect in this case. Normally, to avoid memory leaks, the method should not be called explicitly. It is called implicitly by the matrix assignment operator. The reference counter increment is an atomic operation on the platforms that support it. Thus, it is safe to operate on the same matrices asynchronously in different threads.
1435 Decrements the reference counter and deallocates the matrix if needed.
1437 .. ocv:function:: void Mat::release()
1439 The method decrements the reference counter associated with the matrix data. When the reference counter reaches 0, the matrix data is deallocated and the data and the reference counter pointers are set to NULL's. If the matrix header points to an external data set (see
1440 :ocv:func:`Mat::Mat` ), the reference counter is NULL, and the method has no effect in this case.
1442 This method can be called manually to force the matrix data deallocation. But since this method is automatically called in the destructor, or by any other method that changes the data pointer, it is usually not needed. The reference counter decrement and check for 0 is an atomic operation on the platforms that support it. Thus, it is safe to operate on the same matrices asynchronously in different threads.
1446 Changes the number of matrix rows.
1448 .. ocv:function:: void Mat::resize( size_t sz )
1449 .. ocv:function:: void Mat::resize( size_t sz, const Scalar& s )
1451 :param sz: New number of rows.
1452 :param s: Value assigned to the newly added elements.
1454 The methods change the number of matrix rows. If the matrix is reallocated, the first ``min(Mat::rows, sz)`` rows are preserved. The methods emulate the corresponding methods of the STL vector class.
1459 Reserves space for the certain number of rows.
1461 .. ocv:function:: void Mat::reserve( size_t sz )
1463 :param sz: Number of rows.
1465 The method reserves space for ``sz`` rows. If the matrix already has enough space to store ``sz`` rows, nothing happens. If the matrix is reallocated, the first ``Mat::rows`` rows are preserved. The method emulates the corresponding method of the STL vector class.
1469 Adds elements to the bottom of the matrix.
1471 .. ocv:function:: template<typename T> void Mat::push_back(const T& elem)
1473 .. ocv:function:: void Mat::push_back( const Mat& m )
1475 :param elem: Added element(s).
1476 :param m: Added line(s).
1478 The methods add one or more elements to the bottom of the matrix. They emulate the corresponding method of the STL vector class. When ``elem`` is ``Mat`` , its type and the number of columns must be the same as in the container matrix.
1482 Removes elements from the bottom of the matrix.
1484 .. ocv:function:: template<typename T> void Mat::pop_back(size_t nelems=1)
1486 :param nelems: Number of removed rows. If it is greater than the total number of rows, an exception is thrown.
1488 The method removes one or more rows from the bottom of the matrix.
1493 Locates the matrix header within a parent matrix.
1495 .. ocv:function:: void Mat::locateROI( Size& wholeSize, Point& ofs ) const
1497 :param wholeSize: Output parameter that contains the size of the whole matrix containing ``*this`` as a part.
1499 :param ofs: Output parameter that contains an offset of ``*this`` inside the whole matrix.
1501 After you extracted a submatrix from a matrix using
1502 :ocv:func:`Mat::row`,
1503 :ocv:func:`Mat::col`,
1504 :ocv:func:`Mat::rowRange`,
1505 :ocv:func:`Mat::colRange` , and others, the resultant submatrix points just to the part of the original big matrix. However, each submatrix contains information (represented by ``datastart`` and ``dataend`` fields) that helps reconstruct the original matrix size and the position of the extracted submatrix within the original matrix. The method ``locateROI`` does exactly that.
1510 Adjusts a submatrix size and position within the parent matrix.
1512 .. ocv:function:: Mat& Mat::adjustROI( int dtop, int dbottom, int dleft, int dright )
1514 :param dtop: Shift of the top submatrix boundary upwards.
1516 :param dbottom: Shift of the bottom submatrix boundary downwards.
1518 :param dleft: Shift of the left submatrix boundary to the left.
1520 :param dright: Shift of the right submatrix boundary to the right.
1522 The method is complimentary to
1523 :ocv:func:`Mat::locateROI` . The typical use of these functions is to determine the submatrix position within the parent matrix and then shift the position somehow. Typically, it can be required for filtering operations when pixels outside of the ROI should be taken into account. When all the method parameters are positive, the ROI needs to grow in all directions by the specified amount, for example: ::
1525 A.adjustROI(2, 2, 2, 2);
1528 In this example, the matrix size is increased by 4 elements in each direction. The matrix is shifted by 2 elements to the left and 2 elements up, which brings in all the necessary pixels for the filtering with the 5x5 kernel.
1530 ``adjustROI`` forces the adjusted ROI to be inside of the parent matrix that is boundaries of the adjusted ROI are constrained by boundaries of the parent matrix. For example, if the submatrix ``A`` is located in the first row of a parent matrix and you called ``A.adjustROI(2, 2, 2, 2)`` then ``A`` will not be increased in the upward direction.
1532 The function is used internally by the OpenCV filtering functions, like
1533 :ocv:func:`filter2D` , morphological operations, and so on.
1535 .. seealso:: :ocv:func:`copyMakeBorder`
1540 Extracts a rectangular submatrix.
1542 .. ocv:function:: Mat Mat::operator()( Range rowRange, Range colRange ) const
1544 .. ocv:function:: Mat Mat::operator()( const Rect& roi ) const
1546 .. ocv:function:: Mat Mat::operator()( const Range* ranges ) const
1549 :param rowRange: Start and end row of the extracted submatrix. The upper boundary is not included. To select all the rows, use ``Range::all()``.
1551 :param colRange: Start and end column of the extracted submatrix. The upper boundary is not included. To select all the columns, use ``Range::all()``.
1553 :param roi: Extracted submatrix specified as a rectangle.
1555 :param ranges: Array of selected ranges along each array dimension.
1557 The operators make a new header for the specified sub-array of ``*this`` . They are the most generalized forms of
1558 :ocv:func:`Mat::row`,
1559 :ocv:func:`Mat::col`,
1560 :ocv:func:`Mat::rowRange`, and
1561 :ocv:func:`Mat::colRange` . For example, ``A(Range(0, 10), Range::all())`` is equivalent to ``A.rowRange(0, 10)`` . Similarly to all of the above, the operators are O(1) operations, that is, no matrix data is copied.
1566 Creates the ``CvMat`` header for the matrix.
1568 .. ocv:function:: Mat::operator CvMat() const
1571 The operator creates the ``CvMat`` header for the matrix without copying the underlying data. The reference counter is not taken into account by this operation. Thus, you should make sure than the original matrix is not deallocated while the ``CvMat`` header is used. The operator is useful for intermixing the new and the old OpenCV API's, for example: ::
1573 Mat img(Size(320, 240), CV_8UC3);
1577 mycvOldFunc( &cvimg, ...);
1580 where ``mycvOldFunc`` is a function written to work with OpenCV 1.x data structures.
1583 Mat::operator IplImage
1584 ----------------------
1585 Creates the ``IplImage`` header for the matrix.
1587 .. ocv:function:: Mat::operator IplImage() const
1589 The operator creates the ``IplImage`` header for the matrix without copying the underlying data. You should make sure than the original matrix is not deallocated while the ``IplImage`` header is used. Similarly to ``Mat::operator CvMat`` , the operator is useful for intermixing the new and the old OpenCV API's.
1593 Returns the total number of array elements.
1595 .. ocv:function:: size_t Mat::total() const
1597 The method returns the number of array elements (a number of pixels if the array represents an image).
1601 Reports whether the matrix is continuous or not.
1603 .. ocv:function:: bool Mat::isContinuous() const
1605 The method returns ``true`` if the matrix elements are stored continuously without gaps at the end of each row. Otherwise, it returns ``false``. Obviously, ``1x1`` or ``1xN`` matrices are always continuous. Matrices created with
1606 :ocv:func:`Mat::create` are always continuous. But if you extract a part of the matrix using
1607 :ocv:func:`Mat::col`,
1608 :ocv:func:`Mat::diag` , and so on, or constructed a matrix header for externally allocated data, such matrices may no longer have this property.
1610 The continuity flag is stored as a bit in the ``Mat::flags`` field and is computed automatically when you construct a matrix header. Thus, the continuity check is a very fast operation, though theoretically it could be done as follows: ::
1612 // alternative implementation of Mat::isContinuous()
1613 bool myCheckMatContinuity(const Mat& m)
1615 //return (m.flags & Mat::CONTINUOUS_FLAG) != 0;
1616 return m.rows == 1 || m.step == m.cols*m.elemSize();
1620 The method is used in quite a few of OpenCV functions. The point is that element-wise operations (such as arithmetic and logical operations, math functions, alpha blending, color space transformations, and others) do not depend on the image geometry. Thus, if all the input and output arrays are continuous, the functions can process them as very long single-row vectors. The example below illustrates how an alpha-blending function can be implemented. ::
1622 template<typename T>
1623 void alphaBlendRGBA(const Mat& src1, const Mat& src2, Mat& dst)
1625 const float alpha_scale = (float)std::numeric_limits<T>::max(),
1626 inv_scale = 1.f/alpha_scale;
1628 CV_Assert( src1.type() == src2.type() &&
1629 src1.type() == CV_MAKETYPE(DataType<T>::depth, 4) &&
1630 src1.size() == src2.size());
1631 Size size = src1.size();
1632 dst.create(size, src1.type());
1634 // here is the idiom: check the arrays for continuity and,
1635 // if this is the case,
1636 // treat the arrays as 1D vectors
1637 if( src1.isContinuous() && src2.isContinuous() && dst.isContinuous() )
1639 size.width *= size.height;
1644 for( int i = 0; i < size.height; i++ )
1646 // when the arrays are continuous,
1647 // the outer loop is executed only once
1648 const T* ptr1 = src1.ptr<T>(i);
1649 const T* ptr2 = src2.ptr<T>(i);
1650 T* dptr = dst.ptr<T>(i);
1652 for( int j = 0; j < size.width; j += 4 )
1654 float alpha = ptr1[j+3]*inv_scale, beta = ptr2[j+3]*inv_scale;
1655 dptr[j] = saturate_cast<T>(ptr1[j]*alpha + ptr2[j]*beta);
1656 dptr[j+1] = saturate_cast<T>(ptr1[j+1]*alpha + ptr2[j+1]*beta);
1657 dptr[j+2] = saturate_cast<T>(ptr1[j+2]*alpha + ptr2[j+2]*beta);
1658 dptr[j+3] = saturate_cast<T>((1 - (1-alpha)*(1-beta))*alpha_scale);
1664 This approach, while being very simple, can boost the performance of a simple element-operation by 10-20 percents, especially if the image is rather small and the operation is quite simple.
1666 Another OpenCV idiom in this function, a call of
1667 :ocv:func:`Mat::create` for the destination array, that allocates the destination array unless it already has the proper size and type. And while the newly allocated arrays are always continuous, you still need to check the destination array because :ocv:func:`Mat::create` does not always allocate a new matrix.
1672 Returns the matrix element size in bytes.
1674 .. ocv:function:: size_t Mat::elemSize() const
1676 The method returns the matrix element size in bytes. For example, if the matrix type is ``CV_16SC3`` , the method returns ``3*sizeof(short)`` or 6.
1681 Returns the size of each matrix element channel in bytes.
1683 .. ocv:function:: size_t Mat::elemSize1() const
1685 The method returns the matrix element channel size in bytes, that is, it ignores the number of channels. For example, if the matrix type is ``CV_16SC3`` , the method returns ``sizeof(short)`` or 2.
1690 Returns the type of a matrix element.
1692 .. ocv:function:: int Mat::type() const
1694 The method returns a matrix element type. This is an identifier compatible with the ``CvMat`` type system, like ``CV_16SC3`` or 16-bit signed 3-channel array, and so on.
1699 Returns the depth of a matrix element.
1701 .. ocv:function:: int Mat::depth() const
1703 The method returns the identifier of the matrix element depth (the type of each individual channel). For example, for a 16-bit signed element array, the method returns ``CV_16S`` . A complete list of matrix types contains the following values:
1705 * ``CV_8U`` - 8-bit unsigned integers ( ``0..255`` )
1707 * ``CV_8S`` - 8-bit signed integers ( ``-128..127`` )
1709 * ``CV_16U`` - 16-bit unsigned integers ( ``0..65535`` )
1711 * ``CV_16S`` - 16-bit signed integers ( ``-32768..32767`` )
1713 * ``CV_32S`` - 32-bit signed integers ( ``-2147483648..2147483647`` )
1715 * ``CV_32F`` - 32-bit floating-point numbers ( ``-FLT_MAX..FLT_MAX, INF, NAN`` )
1717 * ``CV_64F`` - 64-bit floating-point numbers ( ``-DBL_MAX..DBL_MAX, INF, NAN`` )
1722 Returns the number of matrix channels.
1724 .. ocv:function:: int Mat::channels() const
1726 The method returns the number of matrix channels.
1731 Returns a normalized step.
1733 .. ocv:function:: size_t Mat::step1( int i=0 ) const
1735 The method returns a matrix step divided by
1736 :ocv:func:`Mat::elemSize1()` . It can be useful to quickly access an arbitrary matrix element.
1741 Returns a matrix size.
1743 .. ocv:function:: Size Mat::size() const
1745 The method returns a matrix size: ``Size(cols, rows)`` . When the matrix is more than 2-dimensional, the returned size is (-1, -1).
1750 Returns ``true`` if the array has no elements.
1752 .. ocv:function:: bool Mat::empty() const
1754 The method returns ``true`` if ``Mat::total()`` is 0 or if ``Mat::data`` is NULL. Because of ``pop_back()`` and ``resize()`` methods ``M.total() == 0`` does not imply that ``M.data == NULL`` .
1759 Returns a pointer to the specified matrix row.
1761 .. ocv:function:: uchar* Mat::ptr(int i0=0)
1763 .. ocv:function:: const uchar* Mat::ptr(int i0=0) const
1765 .. ocv:function:: template<typename _Tp> _Tp* Mat::ptr(int i0=0)
1767 .. ocv:function:: template<typename _Tp> const _Tp* Mat::ptr(int i0=0) const
1769 :param i0: A 0-based row index.
1771 The methods return ``uchar*`` or typed pointer to the specified matrix row. See the sample in
1772 :ocv:func:`Mat::isContinuous` to know how to use these methods.
1777 Returns a reference to the specified array element.
1779 .. ocv:function:: template<typename T> T& Mat::at(int i) const
1781 .. ocv:function:: template<typename T> const T& Mat::at(int i) const
1783 .. ocv:function:: template<typename T> T& Mat::at(int i, int j)
1785 .. ocv:function:: template<typename T> const T& Mat::at(int i, int j) const
1787 .. ocv:function:: template<typename T> T& Mat::at(Point pt)
1789 .. ocv:function:: template<typename T> const T& Mat::at(Point pt) const
1791 .. ocv:function:: template<typename T> T& Mat::at(int i, int j, int k)
1793 .. ocv:function:: template<typename T> const T& Mat::at(int i, int j, int k) const
1795 .. ocv:function:: template<typename T> T& Mat::at(const int* idx)
1797 .. ocv:function:: template<typename T> const T& Mat::at(const int* idx) const
1799 :param i: Index along the dimension 0
1800 :param j: Index along the dimension 1
1801 :param k: Index along the dimension 2
1803 :param pt: Element position specified as ``Point(j,i)`` .
1805 :param idx: Array of ``Mat::dims`` indices.
1807 The template methods return a reference to the specified array element. For the sake of higher performance, the index range checks are only performed in the Debug configuration.
1809 Note that the variants with a single index (i) can be used to access elements of single-row or single-column 2-dimensional arrays. That is, if, for example, ``A`` is a ``1 x N`` floating-point matrix and ``B`` is an ``M x 1`` integer matrix, you can simply write ``A.at<float>(k+4)`` and ``B.at<int>(2*i+1)`` instead of ``A.at<float>(0,k+4)`` and ``B.at<int>(2*i+1,0)`` , respectively.
1811 The example below initializes a Hilbert matrix: ::
1813 Mat H(100, 100, CV_64F);
1814 for(int i = 0; i < H.rows; i++)
1815 for(int j = 0; j < H.cols; j++)
1816 H.at<double>(i,j)=1./(i+j+1);
1822 Returns the matrix iterator and sets it to the first matrix element.
1824 .. ocv:function:: template<typename _Tp> MatIterator_<_Tp> Mat::begin()
1826 .. ocv:function:: template<typename _Tp> MatConstIterator_<_Tp> Mat::begin() const
1828 The methods return the matrix read-only or read-write iterators. The use of matrix iterators is very similar to the use of bi-directional STL iterators. In the example below, the alpha blending function is rewritten using the matrix iterators: ::
1830 template<typename T>
1831 void alphaBlendRGBA(const Mat& src1, const Mat& src2, Mat& dst)
1833 typedef Vec<T, 4> VT;
1835 const float alpha_scale = (float)std::numeric_limits<T>::max(),
1836 inv_scale = 1.f/alpha_scale;
1838 CV_Assert( src1.type() == src2.type() &&
1839 src1.type() == DataType<VT>::type &&
1840 src1.size() == src2.size());
1841 Size size = src1.size();
1842 dst.create(size, src1.type());
1844 MatConstIterator_<VT> it1 = src1.begin<VT>(), it1_end = src1.end<VT>();
1845 MatConstIterator_<VT> it2 = src2.begin<VT>();
1846 MatIterator_<VT> dst_it = dst.begin<VT>();
1848 for( ; it1 != it1_end; ++it1, ++it2, ++dst_it )
1850 VT pix1 = *it1, pix2 = *it2;
1851 float alpha = pix1[3]*inv_scale, beta = pix2[3]*inv_scale;
1852 *dst_it = VT(saturate_cast<T>(pix1[0]*alpha + pix2[0]*beta),
1853 saturate_cast<T>(pix1[1]*alpha + pix2[1]*beta),
1854 saturate_cast<T>(pix1[2]*alpha + pix2[2]*beta),
1855 saturate_cast<T>((1 - (1-alpha)*(1-beta))*alpha_scale));
1863 Returns the matrix iterator and sets it to the after-last matrix element.
1865 .. ocv:function:: template<typename _Tp> MatIterator_<_Tp> Mat::end()
1867 .. ocv:function:: template<typename _Tp> MatConstIterator_<_Tp> Mat::end() const
1869 The methods return the matrix read-only or read-write iterators, set to the point following the last matrix element.
1875 Template matrix class derived from
1876 :ocv:class:`Mat` . ::
1878 template<typename _Tp> class Mat_ : public Mat
1881 // ... some specific methods
1883 // no new extra fields
1887 The class ``Mat_<_Tp>`` is a "thin" template wrapper on top of the ``Mat`` class. It does not have any extra data fields. Nor this class nor ``Mat`` has any virtual methods. Thus, references or pointers to these two classes can be freely but carefully converted one to another. For example: ::
1889 // create a 100x100 8-bit matrix
1890 Mat M(100,100,CV_8U);
1891 // this will be compiled fine. no any data conversion will be done.
1892 Mat_<float>& M1 = (Mat_<float>&)M;
1893 // the program is likely to crash at the statement below
1897 While ``Mat`` is sufficient in most cases, ``Mat_`` can be more convenient if you use a lot of element access operations and if you know matrix type at the compilation time. Note that ``Mat::at<_Tp>(int y, int x)`` and ``Mat_<_Tp>::operator ()(int y, int x)`` do absolutely the same and run at the same speed, but the latter is certainly shorter: ::
1899 Mat_<double> M(20,20);
1900 for(int i = 0; i < M.rows; i++)
1901 for(int j = 0; j < M.cols; j++)
1902 M(i,j) = 1./(i+j+1);
1905 cout << E.at<double>(0,0)/E.at<double>(M.rows-1,0);
1908 To use ``Mat_`` for multi-channel images/matrices, pass ``Vec`` as a ``Mat_`` parameter: ::
1910 // allocate a 320x240 color image and fill it with green (in RGB space)
1911 Mat_<Vec3b> img(240, 320, Vec3b(0,255,0));
1912 // now draw a diagonal white line
1913 for(int i = 0; i < 100; i++)
1914 img(i,i)=Vec3b(255,255,255);
1915 // and now scramble the 2nd (red) channel of each pixel
1916 for(int i = 0; i < img.rows; i++)
1917 for(int j = 0; j < img.cols; j++)
1918 img(i,j)[2] ^= (uchar)(i ^ j);
1923 .. ocv:class:: InputArray
1925 This is the proxy class for passing read-only input arrays into OpenCV functions. It is defined as ::
1927 typedef const _InputArray& InputArray;
1929 where ``_InputArray`` is a class that can be constructed from ``Mat``, ``Mat_<T>``, ``Matx<T, m, n>``, ``std::vector<T>``, ``std::vector<std::vector<T> >`` or ``std::vector<Mat>``. It can also be constructed from a matrix expression.
1931 Since this is mostly implementation-level class, and its interface may change in future versions, we do not describe it in details. There are a few key things, though, that should be kept in mind:
1933 * When you see in the reference manual or in OpenCV source code a function that takes ``InputArray``, it means that you can actually pass ``Mat``, ``Matx``, ``vector<T>`` etc. (see above the complete list).
1935 * Optional input arguments: If some of the input arrays may be empty, pass ``cv::noArray()`` (or simply ``cv::Mat()`` as you probably did before).
1937 * The class is designed solely for passing parameters. That is, normally you *should not* declare class members, local and global variables of this type.
1939 * If you want to design your own function or a class method that can operate of arrays of multiple types, you can use ``InputArray`` (or ``OutputArray``) for the respective parameters. Inside a function you should use ``_InputArray::getMat()`` method to construct a matrix header for the array (without copying data). ``_InputArray::kind()`` can be used to distinguish ``Mat`` from ``vector<>`` etc., but normally it is not needed.
1941 Here is how you can use a function that takes ``InputArray`` ::
1943 std::vector<Point2f> vec;
1944 // points or a circle
1945 for( int i = 0; i < 30; i++ )
1946 vec.push_back(Point2f((float)(100 + 30*cos(i*CV_PI*2/5)),
1947 (float)(100 - 30*sin(i*CV_PI*2/5))));
1948 cv::transform(vec, vec, cv::Matx23f(0.707, -0.707, 10, 0.707, 0.707, 20));
1950 That is, we form an STL vector containing points, and apply in-place affine transformation to the vector using the 2x3 matrix created inline as ``Matx<float, 2, 3>`` instance.
1952 Here is how such a function can be implemented (for simplicity, we implement a very specific case of it, according to the assertion statement inside) ::
1954 void myAffineTransform(InputArray _src, OutputArray _dst, InputArray _m)
1956 // get Mat headers for input arrays. This is O(1) operation,
1957 // unless _src and/or _m are matrix expressions.
1958 Mat src = _src.getMat(), m = _m.getMat();
1959 CV_Assert( src.type() == CV_32FC2 && m.type() == CV_32F && m.size() == Size(3, 2) );
1961 // [re]create the output array so that it has the proper size and type.
1962 // In case of Mat it calls Mat::create, in case of STL vector it calls vector::resize.
1963 _dst.create(src.size(), src.type());
1964 Mat dst = _dst.getMat();
1966 for( int i = 0; i < src.rows; i++ )
1967 for( int j = 0; j < src.cols; j++ )
1969 Point2f pt = src.at<Point2f>(i, j);
1970 dst.at<Point2f>(i, j) = Point2f(m.at<float>(0, 0)*pt.x +
1971 m.at<float>(0, 1)*pt.y +
1973 m.at<float>(1, 0)*pt.x +
1974 m.at<float>(1, 1)*pt.y +
1979 There is another related type, ``InputArrayOfArrays``, which is currently defined as a synonym for ``InputArray``: ::
1981 typedef InputArray InputArrayOfArrays;
1983 It denotes function arguments that are either vectors of vectors or vectors of matrices. A separate synonym is needed to generate Python/Java etc. wrappers properly. At the function implementation level their use is similar, but ``_InputArray::getMat(idx)`` should be used to get header for the idx-th component of the outer vector and ``_InputArray::size().area()`` should be used to find the number of components (vectors/matrices) of the outer vector.
1988 .. ocv:class:: OutputArray : public InputArray
1990 This type is very similar to ``InputArray`` except that it is used for input/output and output function parameters. Just like with ``InputArray``, OpenCV users should not care about ``OutputArray``, they just pass ``Mat``, ``vector<T>`` etc. to the functions. The same limitation as for ``InputArray``: **Do not explicitly create OutputArray instances** applies here too.
1992 If you want to make your function polymorphic (i.e. accept different arrays as output parameters), it is also not very difficult. Take the sample above as the reference. Note that ``_OutputArray::create()`` needs to be called before ``_OutputArray::getMat()``. This way you guarantee that the output array is properly allocated.
1994 Optional output parameters. If you do not need certain output array to be computed and returned to you, pass ``cv::noArray()``, just like you would in the case of optional input array. At the implementation level, use ``_OutputArray::needed()`` to check if certain output array needs to be computed or not.
1996 There are several synonyms for ``OutputArray`` that are used to assist automatic Python/Java/... wrapper generators: ::
1998 typedef OutputArray OutputArrayOfArrays;
1999 typedef OutputArray InputOutputArray;
2000 typedef OutputArray InputOutputArrayOfArrays;
2004 .. ocv:class:: NAryMatIterator
2006 n-ary multi-dimensional array iterator. ::
2008 class CV_EXPORTS NAryMatIterator
2011 //! the default constructor
2013 //! the full constructor taking arbitrary number of n-dim matrices
2014 NAryMatIterator(const Mat** arrays, Mat* planes, int narrays=-1);
2015 //! the separate iterator initialization method
2016 void init(const Mat** arrays, Mat* planes, int narrays=-1);
2018 //! proceeds to the next plane of every iterated matrix
2019 NAryMatIterator& operator ++();
2020 //! proceeds to the next plane of every iterated matrix (postfix increment operator)
2021 NAryMatIterator operator ++(int);
2024 int nplanes; // the total number of planes
2028 Use the class to implement unary, binary, and, generally, n-ary element-wise operations on multi-dimensional arrays. Some of the arguments of an n-ary function may be continuous arrays, some may be not. It is possible to use conventional
2029 ``MatIterator`` 's for each array but incrementing all of the iterators after each small operations may be a big overhead. In this case consider using ``NAryMatIterator`` to iterate through several matrices simultaneously as long as they have the same geometry (dimensionality and all the dimension sizes are the same). On each iteration ``it.planes[0]``, ``it.planes[1]`` , ... will be the slices of the corresponding matrices.
2031 The example below illustrates how you can compute a normalized and threshold 3D color histogram: ::
2033 void computeNormalizedColorHist(const Mat& image, Mat& hist, int N, double minProb)
2035 const int histSize[] = {N, N, N};
2037 // make sure that the histogram has a proper size and type
2038 hist.create(3, histSize, CV_32F);
2043 // the loop below assumes that the image
2044 // is a 8-bit 3-channel. check it.
2045 CV_Assert(image.type() == CV_8UC3);
2046 MatConstIterator_<Vec3b> it = image.begin<Vec3b>(),
2047 it_end = image.end<Vec3b>();
2048 for( ; it != it_end; ++it )
2050 const Vec3b& pix = *it;
2051 hist.at<float>(pix[0]*N/256, pix[1]*N/256, pix[2]*N/256) += 1.f;
2054 minProb *= image.rows*image.cols;
2056 NAryMatIterator it(&hist, &plane, 1);
2058 // iterate through the matrix. on each iteration
2059 // it.planes[*] (of type Mat) will be set to the current plane.
2060 for(int p = 0; p < it.nplanes; p++, ++it)
2062 threshold(it.planes[0], it.planes[0], minProb, 0, THRESH_TOZERO);
2063 s += sum(it.planes[0])[0];
2067 it = NAryMatIterator(&hist, &plane, 1);
2068 for(int p = 0; p < it.nplanes; p++, ++it)
2075 .. ocv:class:: SparseMat
2077 The class ``SparseMat`` represents multi-dimensional sparse numerical arrays. Such a sparse array can store elements of any type that
2078 :ocv:class:`Mat` can store. *Sparse* means that only non-zero elements are stored (though, as a result of operations on a sparse matrix, some of its stored elements can actually become 0. It is up to you to detect such elements and delete them using ``SparseMat::erase`` ). The non-zero elements are stored in a hash table that grows when it is filled so that the search time is O(1) in average (regardless of whether element is there or not). Elements can be accessed using the following methods:
2081 Query operations (``SparseMat::ptr`` and the higher-level ``SparseMat::ref``, ``SparseMat::value`` and ``SparseMat::find``), for example:
2086 int size[] = {10, 10, 10, 10, 10};
2087 SparseMat sparse_mat(dims, size, CV_32F);
2088 for(int i = 0; i < 1000; i++)
2091 for(int k = 0; k < dims; k++)
2093 sparse_mat.ref<float>(idx) += 1.f;
2099 Sparse matrix iterators. They are similar to ``MatIterator`` but different from :ocv:class:`NAryMatIterator`. That is, the iteration loop is familiar to STL users:
2103 // prints elements of a sparse floating-point matrix
2104 // and the sum of elements.
2105 SparseMatConstIterator_<float>
2106 it = sparse_mat.begin<float>(),
2107 it_end = sparse_mat.end<float>();
2109 int dims = sparse_mat.dims();
2110 for(; it != it_end; ++it)
2112 // print element indices and the element value
2113 const SparseMat::Node* n = it.node();
2115 for(int i = 0; i < dims; i++)
2116 printf("%d%s", n->idx[i], i < dims-1 ? ", " : ")");
2117 printf(": %g\n", it.value<float>());
2120 printf("Element sum is %g\n", s);
2124 If you run this loop, you will notice that elements are not enumerated in a logical order (lexicographical, and so on). They come in the same order as they are stored in the hash table (semi-randomly). You may collect pointers to the nodes and sort them to get the proper ordering. Note, however, that pointers to the nodes may become invalid when you add more elements to the matrix. This may happen due to possible buffer reallocation.
2127 Combination of the above 2 methods when you need to process 2 or more sparse matrices simultaneously. For example, this is how you can compute unnormalized cross-correlation of the 2 floating-point sparse matrices:
2131 double cross_corr(const SparseMat& a, const SparseMat& b)
2133 const SparseMat *_a = &a, *_b = &b;
2134 // if b contains less elements than a,
2135 // it is faster to iterate through b
2136 if(_a->nzcount() > _b->nzcount())
2138 SparseMatConstIterator_<float> it = _a->begin<float>(),
2139 it_end = _a->end<float>();
2141 for(; it != it_end; ++it)
2143 // take the next element from the first matrix
2145 const Node* anode = it.node();
2146 // and try to find an element with the same index in the second matrix.
2147 // since the hash value depends only on the element index,
2148 // reuse the hash value stored in the node
2149 float bvalue = _b->value<float>(anode->idx,&anode->hashval);
2150 ccorr += avalue*bvalue;
2157 SparseMat::SparseMat
2158 --------------------
2159 Various SparseMat constructors.
2161 .. ocv:function:: SparseMat::SparseMat()
2162 .. ocv:function:: SparseMat::SparseMat( int dims, const int* _sizes, int _type )
2163 .. ocv:function:: SparseMat::SparseMat( const SparseMat& m )
2164 .. ocv:function:: SparseMat::SparseMat( const Mat& m )
2165 .. ocv:function:: SparseMat::SparseMat( const CvSparseMat* m )
2168 :param m: Source matrix for copy constructor. If m is dense matrix (ocv:class:`Mat`) then it will be converted to sparse representation.
2169 :param dims: Array dimensionality.
2170 :param _sizes: Sparce matrix size on all dementions.
2171 :param _type: Sparse matrix data type.
2173 SparseMat::~SparseMat
2174 ---------------------
2175 SparseMat object destructor.
2177 .. ocv:function:: SparseMat::~SparseMat()
2179 SparseMat::operator=
2180 --------------------
2181 Provides sparse matrix assignment operators.
2183 .. ocv:function:: SparseMat& SparseMat::operator = (const SparseMat& m)
2184 .. ocv:function:: SparseMat& SparseMat::operator = (const Mat& m)
2186 :param m: Matrix for assignment.
2188 The last variant is equivalent to the corresponding constructor with try1d=false.
2193 Creates a full copy of the matrix.
2195 .. ocv:function:: SparseMat SparseMat::clone() const
2199 Copy all the data to the destination matrix.The destination will be reallocated if needed.
2201 .. ocv:function:: void SparseMat::copyTo( SparseMat& m ) const
2202 .. ocv:function:: void SparseMat::copyTo( Mat& m ) const
2204 :param m: Target for copiing.
2206 The last variant converts 1D or 2D sparse matrix to dense 2D matrix. If the sparse matrix is 1D, the result will be a single-column matrix.
2208 SparceMat::convertTo
2209 --------------------
2210 Convert sparse matrix with possible type change and scaling.
2212 .. ocv:function:: void SparseMat::convertTo( SparseMat& m, int rtype, double alpha=1 ) const
2213 .. ocv:function:: void SparseMat::convertTo( Mat& m, int rtype, double alpha=1, double beta=0 ) const
2215 :param m: Destination matrix.
2216 :param rtype: Destination matrix type.
2217 :param alpha: Conversion multiplier.
2219 The first version converts arbitrary sparse matrix to dense matrix and multiplies all the matrix elements by the specified scalar.
2220 The second versiob converts sparse matrix to dense matrix with optional type conversion and scaling.
2221 When rtype=-1, the destination element type will be the same as the sparse matrix element type.
2222 Otherwise, rtype will specify the depth and the number of channels will remain the same as in the sparse matrix.
2226 Reallocates sparse matrix. If it was already of the proper size and type, it is simply cleared with clear(), otherwise,
2227 the old matrix is released (using release()) and the new one is allocated.
2229 .. ocv:function:: void SparseMat::create(int dims, const int* _sizes, int _type)
2231 :param dims: Array dimensionality.
2232 :param _sizes: Sparce matrix size on all dementions.
2233 :param _type: Sparse matrix data type.
2237 Sets all the matrix elements to 0, which means clearing the hash table.
2239 .. ocv:function:: void SparseMat::clear()
2243 Manually increases reference counter to the header.
2245 .. ocv:function:: void SparseMat::addref()
2249 Decreses the header reference counter when it reaches 0. The header and all the underlying data are deallocated.
2251 .. ocv:function:: void SparseMat::release()
2253 SparseMat::CvSparseMat *
2254 ------------------------
2255 Converts sparse matrix to the old-style representation. All the elements are copied.
2257 .. ocv:function:: SparseMat::operator CvSparseMat*() const
2261 Size of each element in bytes (the matrix nodes will be bigger because of element indices and other SparseMat::Node elements).
2263 .. ocv:function:: size_t SparseMat::elemSize() const
2265 SparseMat::elemSize1
2266 --------------------
2267 elemSize()/channels().
2269 .. ocv:function:: size_t SparseMat::elemSize() const
2273 Returns the type of a matrix element.
2275 .. ocv:function:: int SparseMat::type() const
2277 The method returns a sparse matrix element type. This is an identifier compatible with the ``CvMat`` type system, like ``CV_16SC3`` or 16-bit signed 3-channel array, and so on.
2281 Returns the depth of a sparse matrix element.
2283 .. ocv:function:: int SparseMat::depth() const
2285 The method returns the identifier of the matrix element depth (the type of each individual channel). For example, for a 16-bit signed 3-channel array, the method returns ``CV_16S``
2287 * ``CV_8U`` - 8-bit unsigned integers ( ``0..255`` )
2289 * ``CV_8S`` - 8-bit signed integers ( ``-128..127`` )
2291 * ``CV_16U`` - 16-bit unsigned integers ( ``0..65535`` )
2293 * ``CV_16S`` - 16-bit signed integers ( ``-32768..32767`` )
2295 * ``CV_32S`` - 32-bit signed integers ( ``-2147483648..2147483647`` )
2297 * ``CV_32F`` - 32-bit floating-point numbers ( ``-FLT_MAX..FLT_MAX, INF, NAN`` )
2299 * ``CV_64F`` - 64-bit floating-point numbers ( ``-DBL_MAX..DBL_MAX, INF, NAN`` )
2303 Returns the number of matrix channels.
2305 .. ocv:function:: int SparseMat::channels() const
2307 The method returns the number of matrix channels.
2311 Returns the array of sizes or matrix size by i dimension and 0 if the matrix is not allocated.
2313 .. ocv:function:: const int* SparseMat::size() const
2314 .. ocv:function:: int SparseMat::size(int i) const
2316 :param i: Dimention index.
2320 Returns the matrix dimensionality.
2322 .. ocv:function:: int SparseMat::dims() const
2326 Returns the number of non-zero elements.
2328 .. ocv:function:: size_t SparseMat::nzcount() const
2332 Compute element hash value from the element indices.
2334 .. ocv:function:: size_t SparseMat::hash(int i0) const
2335 .. ocv:function:: size_t SparseMat::hash(int i0, int i1) const
2336 .. ocv:function:: size_t SparseMat::hash(int i0, int i1, int i2) const
2337 .. ocv:function:: size_t SparseMat::hash(const int* idx) const
2339 :param i0: The first dimension index.
2340 :param i1: The second dimension index.
2341 :param i2: The third dimension index.
2342 :param idx: Array of element indices for multidimensional matices.
2346 Low-level element-access functions, special variants for 1D, 2D, 3D cases, and the generic one for n-D case.
2348 .. ocv:function:: uchar* SparseMat::ptr(int i0, bool createMissing, size_t* hashval=0)
2349 .. ocv:function:: uchar* SparseMat::ptr(int i0, int i1, bool createMissing, size_t* hashval=0)
2350 .. ocv:function:: uchar* SparseMat::ptr(int i0, int i1, int i2, bool createMissing, size_t* hashval=0)
2351 .. ocv:function:: uchar* SparseMat::ptr(const int* idx, bool createMissing, size_t* hashval=0)
2353 :param i0: The first dimension index.
2354 :param i1: The second dimension index.
2355 :param i2: The third dimension index.
2356 :param idx: Array of element indices for multidimensional matices.
2357 :param createMissing: Create new element with 0 value if it does not exist in SparseMat.
2359 Return pointer to the matrix element. If the element is there (it is non-zero), the pointer to it is returned.
2360 If it is not there and ``createMissing=false``, NULL pointer is returned. If it is not there and ``createMissing=true``,
2361 the new elementis created and initialized with 0. Pointer to it is returned. If the optional hashval pointer is not ``NULL``,
2362 the element hash value is not computed but ``hashval`` is taken instead.
2366 Erase the specified matrix element. When there is no such an element, the methods do nothing.
2368 .. ocv:function:: void SparseMat::erase(int i0, int i1, size_t* hashval=0)
2369 .. ocv:function:: void SparseMat::erase(int i0, int i1, int i2, size_t* hashval=0)
2370 .. ocv:function:: void SparseMat::erase(const int* idx, size_t* hashval=0)
2372 :param i0: The first dimension index.
2373 :param i1: The second dimension index.
2374 :param i2: The third dimension index.
2375 :param idx: Array of element indices for multidimensional matices.
2379 .. ocv:class:: SparseMat_
2381 Template sparse n-dimensional array class derived from
2382 :ocv:class:`SparseMat` ::
2384 template<typename _Tp> class SparseMat_ : public SparseMat
2387 typedef SparseMatIterator_<_Tp> iterator;
2388 typedef SparseMatConstIterator_<_Tp> const_iterator;
2391 // the created matrix will have data type = DataType<_Tp>::type
2393 SparseMat_(int dims, const int* _sizes);
2394 SparseMat_(const SparseMat& m);
2395 SparseMat_(const SparseMat_& m);
2396 SparseMat_(const Mat& m);
2397 SparseMat_(const CvSparseMat* m);
2398 // assignment operators; data type conversion is done when necessary
2399 SparseMat_& operator = (const SparseMat& m);
2400 SparseMat_& operator = (const SparseMat_& m);
2401 SparseMat_& operator = (const Mat& m);
2403 // equivalent to the correspoding parent class methods
2404 SparseMat_ clone() const;
2405 void create(int dims, const int* _sizes);
2406 operator CvSparseMat*() const;
2408 // overriden methods that do extra checks for the data type
2411 int channels() const;
2413 // more convenient element access operations.
2414 // ref() is retained (but <_Tp> specification is not needed anymore);
2415 // operator () is equivalent to SparseMat::value<_Tp>
2416 _Tp& ref(int i0, size_t* hashval=0);
2417 _Tp operator()(int i0, size_t* hashval=0) const;
2418 _Tp& ref(int i0, int i1, size_t* hashval=0);
2419 _Tp operator()(int i0, int i1, size_t* hashval=0) const;
2420 _Tp& ref(int i0, int i1, int i2, size_t* hashval=0);
2421 _Tp operator()(int i0, int i1, int i2, size_t* hashval=0) const;
2422 _Tp& ref(const int* idx, size_t* hashval=0);
2423 _Tp operator()(const int* idx, size_t* hashval=0) const;
2426 SparseMatIterator_<_Tp> begin();
2427 SparseMatConstIterator_<_Tp> begin() const;
2428 SparseMatIterator_<_Tp> end();
2429 SparseMatConstIterator_<_Tp> end() const;
2432 ``SparseMat_`` is a thin wrapper on top of :ocv:class:`SparseMat` created in the same way as ``Mat_`` .
2433 It simplifies notation of some operations. ::
2435 int sz[] = {10, 20, 30};
2436 SparseMat_<double> M(3, sz);
2438 M.ref(1, 2, 3) = M(4, 5, 6) + M(7, 8, 9);
2443 .. ocv:class:: Algorithm
2445 This is a base class for all more or less complex algorithms in OpenCV, especially for classes of algorithms, for which there can be multiple implementations. The examples are stereo correspondence (for which there are algorithms like block matching, semi-global block matching, graph-cut etc.), background subtraction (which can be done using mixture-of-gaussians models, codebook-based algorithm etc.), optical flow (block matching, Lucas-Kanade, Horn-Schunck etc.).
2447 The class provides the following features for all derived classes:
2449 * so called "virtual constructor". That is, each Algorithm derivative is registered at program start and you can get the list of registered algorithms and create instance of a particular algorithm by its name (see ``Algorithm::create``). If you plan to add your own algorithms, it is good practice to add a unique prefix to your algorithms to distinguish them from other algorithms.
2451 * setting/retrieving algorithm parameters by name. If you used video capturing functionality from OpenCV highgui module, you are probably familar with ``cvSetCaptureProperty()``, ``cvGetCaptureProperty()``, ``VideoCapture::set()`` and ``VideoCapture::get()``. ``Algorithm`` provides similar method where instead of integer id's you specify the parameter names as text strings. See ``Algorithm::set`` and ``Algorithm::get`` for details.
2453 * reading and writing parameters from/to XML or YAML files. Every Algorithm derivative can store all its parameters and then read them back. There is no need to re-implement it each time.
2455 Here is example of SIFT use in your application via Algorithm interface: ::
2457 #include "opencv2/opencv.hpp"
2458 #include "opencv2/nonfree/nonfree.hpp"
2462 initModule_nonfree(); // to load SURF/SIFT etc.
2464 Ptr<Feature2D> sift = Algorithm::create<Feature2D>("Feature2D.SIFT");
2466 FileStorage fs("sift_params.xml", FileStorage::READ);
2467 if( fs.isOpened() ) // if we have file with parameters, read them
2469 sift->read(fs["sift_params"]);
2472 else // else modify the parameters and store them; user can later edit the file to use different parameters
2474 sift->set("contrastThreshold", 0.01f); // lower the contrast threshold, compared to the default value
2477 WriteStructContext ws(fs, "sift_params", CV_NODE_MAP);
2482 Mat image = imread("myimage.png", 0), descriptors;
2483 vector<KeyPoint> keypoints;
2484 (*sift)(image, noArray(), keypoints, descriptors);
2488 Returns the algorithm name
2490 .. ocv:function:: string Algorithm::name() const
2494 Returns the algorithm parameter
2496 .. ocv:function:: template<typename _Tp> typename ParamType<_Tp>::member_type Algorithm::get(const string& name) const
2498 :param name: The parameter name.
2500 The method returns value of the particular parameter. Since the compiler can not deduce the type of the returned parameter, you should specify it explicitly in angle brackets. Here are the allowed forms of get:
2502 * myalgo.get<int>("param_name")
2503 * myalgo.get<double>("param_name")
2504 * myalgo.get<bool>("param_name")
2505 * myalgo.get<string>("param_name")
2506 * myalgo.get<Mat>("param_name")
2507 * myalgo.get<vector<Mat> >("param_name")
2508 * myalgo.get<Algorithm>("param_name") (it returns Ptr<Algorithm>).
2510 In some cases the actual type of the parameter can be cast to the specified type, e.g. integer parameter can be cast to double, ``bool`` can be cast to ``int``. But "dangerous" transformations (string<->number, double->int, 1x1 Mat<->number, ...) are not performed and the method will throw an exception. In the case of ``Mat`` or ``vector<Mat>`` parameters the method does not clone the matrix data, so do not modify the matrices. Use ``Algorithm::set`` instead - slower, but more safe.
2515 Sets the algorithm parameter
2517 .. ocv:function:: void Algorithm::set(const string& name, int value)
2518 .. ocv:function:: void Algorithm::set(const string& name, double value)
2519 .. ocv:function:: void Algorithm::set(const string& name, bool value)
2520 .. ocv:function:: void Algorithm::set(const string& name, const string& value)
2521 .. ocv:function:: void Algorithm::set(const string& name, const Mat& value)
2522 .. ocv:function:: void Algorithm::set(const string& name, const vector<Mat>& value)
2523 .. ocv:function:: void Algorithm::set(const string& name, const Ptr<Algorithm>& value)
2525 :param name: The parameter name.
2526 :param value: The parameter value.
2528 The method sets value of the particular parameter. Some of the algorithm parameters may be declared as read-only. If you try to set such a parameter, you will get exception with the corresponding error message.
2533 Stores algorithm parameters in a file storage
2535 .. ocv:function:: void Algorithm::write(FileStorage& fs) const
2537 :param fs: File storage.
2539 The method stores all the algorithm parameters (in alphabetic order) to the file storage. The method is virtual. If you define your own Algorithm derivative, your can override the method and store some extra information. However, it's rarely needed. Here are some examples:
2541 * SIFT feature detector (from nonfree module). The class only stores algorithm parameters and no keypoints or their descriptors. Therefore, it's enough to store the algorithm parameters, which is what ``Algorithm::write()`` does. Therefore, there is no dedicated ``SIFT::write()``.
2543 * Background subtractor (from video module). It has the algorithm parameters and also it has the current background model. However, the background model is not stored. First, it's rather big. Then, if you have stored the background model, it would likely become irrelevant on the next run (because of shifted camera, changed background, different lighting etc.). Therefore, ``BackgroundSubtractorMOG`` and ``BackgroundSubtractorMOG2`` also rely on the standard ``Algorithm::write()`` to store just the algorithm parameters.
2545 * Expectation Maximization (from ml module). The algorithm finds mixture of gaussians that approximates user data best of all. In this case the model may be re-used on the next run to test new data against the trained statistical model. So EM needs to store the model. However, since the model is described by a few parameters that are available as read-only algorithm parameters (i.e. they are available via ``EM::get()``), EM also relies on ``Algorithm::write()`` to store both EM parameters and the model (represented by read-only algorithm parameters).
2550 Reads algorithm parameters from a file storage
2552 .. ocv:function:: void Algorithm::read(const FileNode& fn)
2554 :param fn: File node of the file storage.
2556 The method reads all the algorithm parameters from the specified node of a file storage. Similarly to ``Algorithm::write()``, if you implement an algorithm that needs to read some extra data and/or re-compute some internal data, you may override the method.
2560 Returns the list of registered algorithms
2562 .. ocv:function:: void Algorithm::getList(vector<string>& algorithms)
2564 :param algorithms: The output vector of algorithm names.
2566 This static method returns the list of registered algorithms in alphabetical order. Here is how to use it ::
2568 vector<string> algorithms;
2569 Algorithm::getList(algorithms);
2570 cout << "Algorithms: " << algorithms.size() << endl;
2571 for (size_t i=0; i < algorithms.size(); i++)
2572 cout << algorithms[i] << endl;
2577 Creates algorithm instance by name
2579 .. ocv:function:: template<typename _Tp> Ptr<_Tp> Algorithm::create(const string& name)
2581 :param name: The algorithm name, one of the names returned by ``Algorithm::getList()``.
2583 This static method creates a new instance of the specified algorithm. If there is no such algorithm, the method will silently return null pointer (that can be checked by ``Ptr::empty()`` method). Also, you should specify the particular ``Algorithm`` subclass as ``_Tp`` (or simply ``Algorithm`` if you do not know it at that point). ::
2585 Ptr<BackgroundSubtractor> bgfg = Algorithm::create<BackgroundSubtractor>("BackgroundSubtractor.MOG2");
2587 .. note:: This is important note about seemingly mysterious behavior of ``Algorithm::create()`` when it returns NULL while it should not. The reason is simple - ``Algorithm::create()`` resides in OpenCV`s core module and the algorithms are implemented in other modules. If you create algorithms dynamically, C++ linker may decide to throw away the modules where the actual algorithms are implemented, since you do not call any functions from the modules. To avoid this problem, you need to call ``initModule_<modulename>();`` somewhere in the beginning of the program before ``Algorithm::create()``. For example, call ``initModule_nonfree()`` in order to use SURF/SIFT, call ``initModule_ml()`` to use expectation maximization etc.
2589 Creating Own Algorithms
2590 -----------------------
2592 The above methods are usually enough for users. If you want to make your own algorithm, derived from ``Algorithm``, you should basically follow a few conventions and add a little semi-standard piece of code to your class:
2594 * Make a class and specify ``Algorithm`` as its base class.
2595 * The algorithm parameters should be the class members. See ``Algorithm::get()`` for the list of possible types of the parameters.
2596 * Add public virtual method ``AlgorithmInfo* info() const;`` to your class.
2597 * Add constructor function, ``AlgorithmInfo`` instance and implement the ``info()`` method. The simplest way is to take http://code.opencv.org/projects/opencv/repository/revisions/master/entry/modules/ml/src/ml_init.cpp as the reference and modify it according to the list of your parameters.
2598 * Add some public function (e.g. ``initModule_<mymodule>()``) that calls info() of your algorithm and put it into the same source file as ``info()`` implementation. This is to force C++ linker to include this object file into the target application. See ``Algorithm::create()`` for details.