1 /*M///////////////////////////////////////////////////////////////////////////////////////
\r
3 // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
\r
5 // By downloading, copying, installing or using the software you agree to this license.
\r
6 // If you do not agree to this license, do not download, install,
\r
7 // copy or use the software.
\r
10 // License Agreement
\r
11 // For Open Source Computer Vision Library
\r
13 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
\r
14 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
\r
15 // Third party copyrights are property of their respective owners.
\r
17 // Redistribution and use in source and binary forms, with or without modification,
\r
18 // are permitted provided that the following conditions are met:
\r
20 // * Redistribution's of source code must retain the above copyright notice,
\r
21 // this list of conditions and the following disclaimer.
\r
23 // * Redistribution's in binary form must reproduce the above copyright notice,
\r
24 // this list of conditions and the following disclaimer in the documentation
\r
25 // and/or other GpuMaterials provided with the distribution.
\r
27 // * The name of the copyright holders may not be used to endorse or promote products
\r
28 // derived from this software without specific prior written permission.
\r
30 // This software is provided by the copyright holders and contributors "as is" and
\r
31 // any express or implied warranties, including, but not limited to, the implied
\r
32 // warranties of merchantability and fitness for a particular purpose are disclaimed.
\r
33 // In no event shall the Intel Corporation or contributors be liable for any direct,
\r
34 // indirect, incidental, special, exemplary, or consequential damages
\r
35 // (including, but not limited to, procurement of substitute goods or services;
\r
36 // loss of use, data, or profits; or business interruption) however caused
\r
37 // and on any theory of liability, whether in contract, strict liability,
\r
38 // or tort (including negligence or otherwise) arising in any way out of
\r
39 // the use of this software, even if advised of the possibility of such damage.
\r
43 #ifndef __OPENCV_GPU_HPP__
\r
44 #define __OPENCV_GPU_HPP__
\r
47 #include "opencv2/core/core.hpp"
\r
48 #include "opencv2/imgproc/imgproc.hpp"
\r
49 #include "opencv2/objdetect/objdetect.hpp"
\r
50 #include "opencv2/gpu/devmem2d.hpp"
\r
51 #include "opencv2/features2d/features2d.hpp"
\r
57 //////////////////////////////// Initialization & Info ////////////////////////
\r
59 //! This is the only function that do not throw exceptions if the library is compiled without Cuda.
\r
60 CV_EXPORTS int getCudaEnabledDeviceCount();
\r
62 //! Functions below throw cv::Expception if the library is compiled without Cuda.
\r
64 CV_EXPORTS void setDevice(int device);
\r
65 CV_EXPORTS int getDevice();
\r
67 //! Explicitly destroys and cleans up all resources associated with the current device in the current process.
\r
68 //! Any subsequent API call to this device will reinitialize the device.
\r
69 CV_EXPORTS void resetDevice();
\r
73 FEATURE_SET_COMPUTE_10 = 10,
\r
74 FEATURE_SET_COMPUTE_11 = 11,
\r
75 FEATURE_SET_COMPUTE_12 = 12,
\r
76 FEATURE_SET_COMPUTE_13 = 13,
\r
77 FEATURE_SET_COMPUTE_20 = 20,
\r
78 FEATURE_SET_COMPUTE_21 = 21,
\r
79 GLOBAL_ATOMICS = FEATURE_SET_COMPUTE_11,
\r
80 NATIVE_DOUBLE = FEATURE_SET_COMPUTE_13
\r
83 // Gives information about what GPU archs this OpenCV GPU module was
\r
85 class CV_EXPORTS TargetArchs
\r
88 static bool builtWith(FeatureSet feature_set);
\r
89 static bool has(int major, int minor);
\r
90 static bool hasPtx(int major, int minor);
\r
91 static bool hasBin(int major, int minor);
\r
92 static bool hasEqualOrLessPtx(int major, int minor);
\r
93 static bool hasEqualOrGreater(int major, int minor);
\r
94 static bool hasEqualOrGreaterPtx(int major, int minor);
\r
95 static bool hasEqualOrGreaterBin(int major, int minor);
\r
100 // Gives information about the given GPU
\r
101 class CV_EXPORTS DeviceInfo
\r
104 // Creates DeviceInfo object for the current GPU
\r
105 DeviceInfo() : device_id_(getDevice()) { query(); }
\r
107 // Creates DeviceInfo object for the given GPU
\r
108 DeviceInfo(int device_id) : device_id_(device_id) { query(); }
\r
110 string name() const { return name_; }
\r
112 // Return compute capability versions
\r
113 int majorVersion() const { return majorVersion_; }
\r
114 int minorVersion() const { return minorVersion_; }
\r
116 int multiProcessorCount() const { return multi_processor_count_; }
\r
118 size_t freeMemory() const;
\r
119 size_t totalMemory() const;
\r
121 // Checks whether device supports the given feature
\r
122 bool supports(FeatureSet feature_set) const;
\r
124 // Checks whether the GPU module can be run on the given device
\r
125 bool isCompatible() const;
\r
127 int deviceID() const { return device_id_; }
\r
131 void queryMemory(size_t& free_memory, size_t& total_memory) const;
\r
136 int multi_processor_count_;
\r
141 //////////////////////////////// Error handling ////////////////////////
\r
143 CV_EXPORTS void error(const char *error_string, const char *file, const int line, const char *func);
\r
144 CV_EXPORTS void nppError( int err, const char *file, const int line, const char *func);
\r
146 //////////////////////////////// GpuMat ////////////////////////////////
\r
150 //! Smart pointer for GPU memory with reference counting. Its interface is mostly similar with cv::Mat.
\r
151 class CV_EXPORTS GpuMat
\r
154 //! default constructor
\r
156 //! constructs GpuMatrix of the specified size and type (_type is CV_8UC1, CV_64FC3, CV_32SC(12) etc.)
\r
157 GpuMat(int rows, int cols, int type);
\r
158 GpuMat(Size size, int type);
\r
159 //! constucts GpuMatrix and fills it with the specified value _s.
\r
160 GpuMat(int rows, int cols, int type, const Scalar& s);
\r
161 GpuMat(Size size, int type, const Scalar& s);
\r
162 //! copy constructor
\r
163 GpuMat(const GpuMat& m);
\r
165 //! constructor for GpuMatrix headers pointing to user-allocated data
\r
166 GpuMat(int rows, int cols, int type, void* data, size_t step = Mat::AUTO_STEP);
\r
167 GpuMat(Size size, int type, void* data, size_t step = Mat::AUTO_STEP);
\r
169 //! creates a matrix header for a part of the bigger matrix
\r
170 GpuMat(const GpuMat& m, const Range& rowRange, const Range& colRange);
\r
171 GpuMat(const GpuMat& m, const Rect& roi);
\r
173 //! builds GpuMat from Mat. Perfom blocking upload to device.
\r
174 explicit GpuMat (const Mat& m);
\r
176 //! destructor - calls release()
\r
179 //! assignment operators
\r
180 GpuMat& operator = (const GpuMat& m);
\r
181 //! assignment operator. Perfom blocking upload to device.
\r
182 GpuMat& operator = (const Mat& m);
\r
184 //! returns lightweight DevMem2D_ structure for passing to nvcc-compiled code.
\r
185 // Contains just image size, data ptr and step.
\r
186 template <class T> operator DevMem2D_<T>() const;
\r
187 template <class T> operator PtrStep_<T>() const;
\r
189 //! pefroms blocking upload data to GpuMat.
\r
190 void upload(const cv::Mat& m);
\r
193 void upload(const CudaMem& m, Stream& stream);
\r
195 //! downloads data from device to host memory. Blocking calls.
\r
196 operator Mat() const;
\r
197 void download(cv::Mat& m) const;
\r
200 void download(CudaMem& m, Stream& stream) const;
\r
202 //! returns a new GpuMatrix header for the specified row
\r
203 GpuMat row(int y) const;
\r
204 //! returns a new GpuMatrix header for the specified column
\r
205 GpuMat col(int x) const;
\r
206 //! ... for the specified row span
\r
207 GpuMat rowRange(int startrow, int endrow) const;
\r
208 GpuMat rowRange(const Range& r) const;
\r
209 //! ... for the specified column span
\r
210 GpuMat colRange(int startcol, int endcol) const;
\r
211 GpuMat colRange(const Range& r) const;
\r
213 //! returns deep copy of the GpuMatrix, i.e. the data is copied
\r
214 GpuMat clone() const;
\r
215 //! copies the GpuMatrix content to "m".
\r
216 // It calls m.create(this->size(), this->type()).
\r
217 void copyTo( GpuMat& m ) const;
\r
218 //! copies those GpuMatrix elements to "m" that are marked with non-zero mask elements.
\r
219 void copyTo( GpuMat& m, const GpuMat& mask ) const;
\r
220 //! converts GpuMatrix to another datatype with optional scalng. See cvConvertScale.
\r
221 void convertTo( GpuMat& m, int rtype, double alpha=1, double beta=0 ) const;
\r
223 void assignTo( GpuMat& m, int type=-1 ) const;
\r
225 //! sets every GpuMatrix element to s
\r
226 GpuMat& operator = (const Scalar& s);
\r
227 //! sets some of the GpuMatrix elements to s, according to the mask
\r
228 GpuMat& setTo(const Scalar& s, const GpuMat& mask = GpuMat());
\r
229 //! creates alternative GpuMatrix header for the same data, with different
\r
230 // number of channels and/or different number of rows. see cvReshape.
\r
231 GpuMat reshape(int cn, int rows = 0) const;
\r
233 //! allocates new GpuMatrix data unless the GpuMatrix already has specified size and type.
\r
234 // previous data is unreferenced if needed.
\r
235 void create(int rows, int cols, int type);
\r
236 void create(Size size, int type);
\r
237 //! decreases reference counter;
\r
238 // deallocate the data when reference counter reaches 0.
\r
241 //! swaps with other smart pointer
\r
242 void swap(GpuMat& mat);
\r
244 //! locates GpuMatrix header within a parent GpuMatrix. See below
\r
245 void locateROI( Size& wholeSize, Point& ofs ) const;
\r
246 //! moves/resizes the current GpuMatrix ROI inside the parent GpuMatrix.
\r
247 GpuMat& adjustROI( int dtop, int dbottom, int dleft, int dright );
\r
248 //! extracts a rectangular sub-GpuMatrix
\r
249 // (this is a generalized form of row, rowRange etc.)
\r
250 GpuMat operator()( Range rowRange, Range colRange ) const;
\r
251 GpuMat operator()( const Rect& roi ) const;
\r
253 //! returns true iff the GpuMatrix data is continuous
\r
254 // (i.e. when there are no gaps between successive rows).
\r
255 // similar to CV_IS_GpuMat_CONT(cvGpuMat->type)
\r
256 bool isContinuous() const;
\r
257 //! returns element size in bytes,
\r
258 // similar to CV_ELEM_SIZE(cvMat->type)
\r
259 size_t elemSize() const;
\r
260 //! returns the size of element channel in bytes.
\r
261 size_t elemSize1() const;
\r
262 //! returns element type, similar to CV_MAT_TYPE(cvMat->type)
\r
264 //! returns element type, similar to CV_MAT_DEPTH(cvMat->type)
\r
266 //! returns element type, similar to CV_MAT_CN(cvMat->type)
\r
267 int channels() const;
\r
268 //! returns step/elemSize1()
\r
269 size_t step1() const;
\r
270 //! returns GpuMatrix size:
\r
271 // width == number of columns, height == number of rows
\r
273 //! returns true if GpuMatrix data is NULL
\r
274 bool empty() const;
\r
276 //! returns pointer to y-th row
\r
277 uchar* ptr(int y = 0);
\r
278 const uchar* ptr(int y = 0) const;
\r
280 //! template version of the above method
\r
281 template<typename _Tp> _Tp* ptr(int y = 0);
\r
282 template<typename _Tp> const _Tp* ptr(int y = 0) const;
\r
284 //! matrix transposition
\r
287 /*! includes several bit-fields:
\r
288 - the magic signature
\r
291 - number of channels
\r
294 //! the number of rows and columns
\r
296 //! a distance between successive rows in bytes; includes the gap if any
\r
298 //! pointer to the data
\r
301 //! pointer to the reference counter;
\r
302 // when GpuMatrix points to user-allocated data, the pointer is NULL
\r
305 //! helper fields used in locateROI and adjustROI
\r
310 //#define TemplatedGpuMat // experimental now, deprecated to use
\r
311 #ifdef TemplatedGpuMat
\r
312 #include "GpuMat_BetaDeprecated.hpp"
\r
315 //! Creates continuous GPU matrix
\r
316 CV_EXPORTS void createContinuous(int rows, int cols, int type, GpuMat& m);
\r
318 //! Ensures that size of the given matrix is not less than (rows, cols) size
\r
319 //! and matrix type is match specified one too
\r
320 CV_EXPORTS void ensureSizeIsEnough(int rows, int cols, int type, GpuMat& m);
\r
322 //////////////////////////////// CudaMem ////////////////////////////////
\r
323 // CudaMem is limited cv::Mat with page locked memory allocation.
\r
324 // Page locked memory is only needed for async and faster coping to GPU.
\r
325 // It is convertable to cv::Mat header without reference counting
\r
326 // so you can use it with other opencv functions.
\r
328 class CV_EXPORTS CudaMem
\r
331 enum { ALLOC_PAGE_LOCKED = 1, ALLOC_ZEROCOPY = 2, ALLOC_WRITE_COMBINED = 4 };
\r
334 CudaMem(const CudaMem& m);
\r
336 CudaMem(int rows, int cols, int type, int _alloc_type = ALLOC_PAGE_LOCKED);
\r
337 CudaMem(Size size, int type, int alloc_type = ALLOC_PAGE_LOCKED);
\r
340 //! creates from cv::Mat with coping data
\r
341 explicit CudaMem(const Mat& m, int alloc_type = ALLOC_PAGE_LOCKED);
\r
345 CudaMem& operator = (const CudaMem& m);
\r
347 //! returns deep copy of the matrix, i.e. the data is copied
\r
348 CudaMem clone() const;
\r
350 //! allocates new matrix data unless the matrix already has specified size and type.
\r
351 void create(int rows, int cols, int type, int alloc_type = ALLOC_PAGE_LOCKED);
\r
352 void create(Size size, int type, int alloc_type = ALLOC_PAGE_LOCKED);
\r
354 //! decrements reference counter and released memory if needed.
\r
357 //! returns matrix header with disabled reference counting for CudaMem data.
\r
358 Mat createMatHeader() const;
\r
359 operator Mat() const;
\r
361 //! maps host memory into device address space and returns GpuMat header for it. Throws exception if not supported by hardware.
\r
362 GpuMat createGpuMatHeader() const;
\r
363 operator GpuMat() const;
\r
365 //returns if host memory can be mapperd to gpu address space;
\r
366 static bool canMapHostMemory();
\r
368 // Please see cv::Mat for descriptions
\r
369 bool isContinuous() const;
\r
370 size_t elemSize() const;
\r
371 size_t elemSize1() const;
\r
374 int channels() const;
\r
375 size_t step1() const;
\r
377 bool empty() const;
\r
380 // Please see cv::Mat for descriptions
\r
394 //////////////////////////////// CudaStream ////////////////////////////////
\r
395 // Encapculates Cuda Stream. Provides interface for async coping.
\r
396 // Passed to each function that supports async kernel execution.
\r
397 // Reference counting is enabled
\r
399 class CV_EXPORTS Stream
\r
405 Stream(const Stream&);
\r
406 Stream& operator=(const Stream&);
\r
408 bool queryIfComplete();
\r
409 void waitForCompletion();
\r
411 //! downloads asynchronously.
\r
412 // Warning! cv::Mat must point to page locked memory (i.e. to CudaMem data or to its subMat)
\r
413 void enqueueDownload(const GpuMat& src, CudaMem& dst);
\r
414 void enqueueDownload(const GpuMat& src, Mat& dst);
\r
416 //! uploads asynchronously.
\r
417 // Warning! cv::Mat must point to page locked memory (i.e. to CudaMem data or to its ROI)
\r
418 void enqueueUpload(const CudaMem& src, GpuMat& dst);
\r
419 void enqueueUpload(const Mat& src, GpuMat& dst);
\r
421 void enqueueCopy(const GpuMat& src, GpuMat& dst);
\r
423 void enqueueMemSet(GpuMat& src, Scalar val);
\r
424 void enqueueMemSet(GpuMat& src, Scalar val, const GpuMat& mask);
\r
426 // converts matrix type, ex from float to uchar depending on type
\r
427 void enqueueConvert(const GpuMat& src, GpuMat& dst, int type, double a = 1, double b = 0);
\r
429 static Stream& Null();
\r
431 operator bool() const;
\r
440 friend struct StreamAccessor;
\r
442 explicit Stream(Impl* impl);
\r
446 ////////////////////////////// Arithmetics ///////////////////////////////////
\r
448 //! transposes the matrix
\r
449 //! supports matrix with element size = 1, 4 and 8 bytes (CV_8UC1, CV_8UC4, CV_16UC2, CV_32FC1, etc)
\r
450 CV_EXPORTS void transpose(const GpuMat& src1, GpuMat& dst, Stream& stream = Stream::Null());
\r
452 //! reverses the order of the rows, columns or both in a matrix
\r
453 //! supports CV_8UC1, CV_8UC4 types
\r
454 CV_EXPORTS void flip(const GpuMat& a, GpuMat& b, int flipCode, Stream& stream = Stream::Null());
\r
456 //! transforms 8-bit unsigned integers using lookup table: dst(i)=lut(src(i))
\r
457 //! destination array will have the depth type as lut and the same channels number as source
\r
458 //! supports CV_8UC1, CV_8UC3 types
\r
459 CV_EXPORTS void LUT(const GpuMat& src, const Mat& lut, GpuMat& dst, Stream& stream = Stream::Null());
\r
461 //! makes multi-channel array out of several single-channel arrays
\r
462 CV_EXPORTS void merge(const GpuMat* src, size_t n, GpuMat& dst, Stream& stream = Stream::Null());
\r
464 //! makes multi-channel array out of several single-channel arrays
\r
465 CV_EXPORTS void merge(const vector<GpuMat>& src, GpuMat& dst, Stream& stream = Stream::Null());
\r
467 //! copies each plane of a multi-channel array to a dedicated array
\r
468 CV_EXPORTS void split(const GpuMat& src, GpuMat* dst, Stream& stream = Stream::Null());
\r
470 //! copies each plane of a multi-channel array to a dedicated array
\r
471 CV_EXPORTS void split(const GpuMat& src, vector<GpuMat>& dst, Stream& stream = Stream::Null());
\r
473 //! computes magnitude of complex (x(i).re, x(i).im) vector
\r
474 //! supports only CV_32FC2 type
\r
475 CV_EXPORTS void magnitude(const GpuMat& x, GpuMat& magnitude, Stream& stream = Stream::Null());
\r
477 //! computes squared magnitude of complex (x(i).re, x(i).im) vector
\r
478 //! supports only CV_32FC2 type
\r
479 CV_EXPORTS void magnitudeSqr(const GpuMat& x, GpuMat& magnitude, Stream& stream = Stream::Null());
\r
481 //! computes magnitude of each (x(i), y(i)) vector
\r
482 //! supports only floating-point source
\r
483 CV_EXPORTS void magnitude(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, Stream& stream = Stream::Null());
\r
485 //! computes squared magnitude of each (x(i), y(i)) vector
\r
486 //! supports only floating-point source
\r
487 CV_EXPORTS void magnitudeSqr(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, Stream& stream = Stream::Null());
\r
489 //! computes angle (angle(i)) of each (x(i), y(i)) vector
\r
490 //! supports only floating-point source
\r
491 CV_EXPORTS void phase(const GpuMat& x, const GpuMat& y, GpuMat& angle, bool angleInDegrees = false, Stream& stream = Stream::Null());
\r
493 //! converts Cartesian coordinates to polar
\r
494 //! supports only floating-point source
\r
495 CV_EXPORTS void cartToPolar(const GpuMat& x, const GpuMat& y, GpuMat& magnitude, GpuMat& angle, bool angleInDegrees = false, Stream& stream = Stream::Null());
\r
497 //! converts polar coordinates to Cartesian
\r
498 //! supports only floating-point source
\r
499 CV_EXPORTS void polarToCart(const GpuMat& magnitude, const GpuMat& angle, GpuMat& x, GpuMat& y, bool angleInDegrees = false, Stream& stream = Stream::Null());
\r
502 //////////////////////////// Per-element operations ////////////////////////////////////
\r
504 //! adds one matrix to another (c = a + b)
\r
505 //! supports CV_8UC1, CV_8UC4, CV_32SC1, CV_32FC1 types
\r
506 CV_EXPORTS void add(const GpuMat& a, const GpuMat& b, GpuMat& c, Stream& stream = Stream::Null());
\r
507 //! adds scalar to a matrix (c = a + s)
\r
508 //! supports CV_32FC1 and CV_32FC2 type
\r
509 CV_EXPORTS void add(const GpuMat& a, const Scalar& sc, GpuMat& c, Stream& stream = Stream::Null());
\r
511 //! subtracts one matrix from another (c = a - b)
\r
512 //! supports CV_8UC1, CV_8UC4, CV_32SC1, CV_32FC1 types
\r
513 CV_EXPORTS void subtract(const GpuMat& a, const GpuMat& b, GpuMat& c, Stream& stream = Stream::Null());
\r
514 //! subtracts scalar from a matrix (c = a - s)
\r
515 //! supports CV_32FC1 and CV_32FC2 type
\r
516 CV_EXPORTS void subtract(const GpuMat& a, const Scalar& sc, GpuMat& c, Stream& stream = Stream::Null());
\r
518 //! computes element-wise product of the two arrays (c = a * b)
\r
519 //! supports CV_8UC1, CV_8UC4, CV_32SC1, CV_32FC1 types
\r
520 CV_EXPORTS void multiply(const GpuMat& a, const GpuMat& b, GpuMat& c, Stream& stream = Stream::Null());
\r
521 //! multiplies matrix to a scalar (c = a * s)
\r
522 //! supports CV_32FC1 type
\r
523 CV_EXPORTS void multiply(const GpuMat& a, const Scalar& sc, GpuMat& c, Stream& stream = Stream::Null());
\r
525 //! computes element-wise quotient of the two arrays (c = a / b)
\r
526 //! supports CV_8UC1, CV_8UC4, CV_32SC1, CV_32FC1 types
\r
527 CV_EXPORTS void divide(const GpuMat& a, const GpuMat& b, GpuMat& c, Stream& stream = Stream::Null());
\r
528 //! computes element-wise quotient of matrix and scalar (c = a / s)
\r
529 //! supports CV_32FC1 type
\r
530 CV_EXPORTS void divide(const GpuMat& a, const Scalar& sc, GpuMat& c, Stream& stream = Stream::Null());
\r
532 //! computes exponent of each matrix element (b = e**a)
\r
533 //! supports only CV_32FC1 type
\r
534 CV_EXPORTS void exp(const GpuMat& a, GpuMat& b, Stream& stream = Stream::Null());
\r
536 //! computes power of each matrix element:
\r
537 // (dst(i,j) = pow( src(i,j) , power), if src.type() is integer
\r
538 // (dst(i,j) = pow(fabs(src(i,j)), power), otherwise
\r
539 //! supports all, except depth == CV_64F
\r
540 CV_EXPORTS void pow(const GpuMat& src, double power, GpuMat& dst, Stream& stream = Stream::Null());
\r
542 //! computes natural logarithm of absolute value of each matrix element: b = log(abs(a))
\r
543 //! supports only CV_32FC1 type
\r
544 CV_EXPORTS void log(const GpuMat& a, GpuMat& b, Stream& stream = Stream::Null());
\r
546 //! computes element-wise absolute difference of two arrays (c = abs(a - b))
\r
547 //! supports CV_8UC1, CV_8UC4, CV_32SC1, CV_32FC1 types
\r
548 CV_EXPORTS void absdiff(const GpuMat& a, const GpuMat& b, GpuMat& c, Stream& stream = Stream::Null());
\r
549 //! computes element-wise absolute difference of array and scalar (c = abs(a - s))
\r
550 //! supports only CV_32FC1 type
\r
551 CV_EXPORTS void absdiff(const GpuMat& a, const Scalar& s, GpuMat& c, Stream& stream = Stream::Null());
\r
553 //! compares elements of two arrays (c = a <cmpop> b)
\r
554 //! supports CV_8UC4, CV_32FC1 types
\r
555 CV_EXPORTS void compare(const GpuMat& a, const GpuMat& b, GpuMat& c, int cmpop, Stream& stream = Stream::Null());
\r
557 //! performs per-elements bit-wise inversion
\r
558 CV_EXPORTS void bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
\r
560 //! calculates per-element bit-wise disjunction of two arrays
\r
561 CV_EXPORTS void bitwise_or(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
\r
563 //! calculates per-element bit-wise conjunction of two arrays
\r
564 CV_EXPORTS void bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
\r
566 //! calculates per-element bit-wise "exclusive or" operation
\r
567 CV_EXPORTS void bitwise_xor(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask=GpuMat(), Stream& stream = Stream::Null());
\r
569 //! computes per-element minimum of two arrays (dst = min(src1, src2))
\r
570 CV_EXPORTS void min(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& stream = Stream::Null());
\r
572 //! computes per-element minimum of array and scalar (dst = min(src1, src2))
\r
573 CV_EXPORTS void min(const GpuMat& src1, double src2, GpuMat& dst, Stream& stream = Stream::Null());
\r
575 //! computes per-element maximum of two arrays (dst = max(src1, src2))
\r
576 CV_EXPORTS void max(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, Stream& stream = Stream::Null());
\r
578 //! computes per-element maximum of array and scalar (dst = max(src1, src2))
\r
579 CV_EXPORTS void max(const GpuMat& src1, double src2, GpuMat& dst, Stream& stream = Stream::Null());
\r
582 ////////////////////////////// Image processing //////////////////////////////
\r
584 //! DST[x,y] = SRC[xmap[x,y],ymap[x,y]] with bilinear interpolation.
\r
585 //! supports CV_8UC1, CV_8UC3 source types and CV_32FC1 map type
\r
586 CV_EXPORTS void remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap);
\r
588 //! Does mean shift filtering on GPU.
\r
589 CV_EXPORTS void meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr,
\r
590 TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1));
\r
592 //! Does mean shift procedure on GPU.
\r
593 CV_EXPORTS void meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr,
\r
594 TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1));
\r
596 //! Does mean shift segmentation with elimination of small regions.
\r
597 CV_EXPORTS void meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr, int minsize,
\r
598 TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1));
\r
600 //! Does coloring of disparity image: [0..ndisp) -> [0..240, 1, 1] in HSV.
\r
601 //! Supported types of input disparity: CV_8U, CV_16S.
\r
602 //! Output disparity has CV_8UC4 type in BGRA format (alpha = 255).
\r
603 CV_EXPORTS void drawColorDisp(const GpuMat& src_disp, GpuMat& dst_disp, int ndisp, Stream& stream = Stream::Null());
\r
605 //! Reprojects disparity image to 3D space.
\r
606 //! Supports CV_8U and CV_16S types of input disparity.
\r
607 //! The output is a 4-channel floating-point (CV_32FC4) matrix.
\r
608 //! Each element of this matrix will contain the 3D coordinates of the point (x,y,z,1), computed from the disparity map.
\r
609 //! Q is the 4x4 perspective transformation matrix that can be obtained with cvStereoRectify.
\r
610 CV_EXPORTS void reprojectImageTo3D(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, Stream& stream = Stream::Null());
\r
612 //! converts image from one color space to another
\r
613 CV_EXPORTS void cvtColor(const GpuMat& src, GpuMat& dst, int code, int dcn = 0, Stream& stream = Stream::Null());
\r
615 //! applies fixed threshold to the image
\r
616 CV_EXPORTS double threshold(const GpuMat& src, GpuMat& dst, double thresh, double maxval, int type, Stream& stream = Stream::Null());
\r
618 //! resizes the image
\r
619 //! Supports INTER_NEAREST, INTER_LINEAR
\r
620 //! supports CV_8UC1, CV_8UC4 types
\r
621 CV_EXPORTS void resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
\r
623 //! warps the image using affine transformation
\r
624 //! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
\r
625 CV_EXPORTS void warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags = INTER_LINEAR, Stream& stream = Stream::Null());
\r
627 //! warps the image using perspective transformation
\r
628 //! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
\r
629 CV_EXPORTS void warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags = INTER_LINEAR, Stream& stream = Stream::Null());
\r
631 //! builds plane warping maps
\r
632 CV_EXPORTS void buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s, double dist,
\r
633 GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
\r
635 //! builds cylindrical warping maps
\r
636 CV_EXPORTS void buildWarpCylindricalMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
\r
637 GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
\r
639 //! builds spherical warping maps
\r
640 CV_EXPORTS void buildWarpSphericalMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
\r
641 GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
\r
643 //! rotate 8bit single or four channel image
\r
644 //! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
\r
645 //! supports CV_8UC1, CV_8UC4 types
\r
646 CV_EXPORTS void rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift = 0, double yShift = 0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
\r
648 //! copies 2D array to a larger destination array and pads borders with user-specifiable constant
\r
649 //! supports CV_8UC1, CV_8UC4, CV_32SC1 and CV_32FC1 types
\r
650 CV_EXPORTS void copyMakeBorder(const GpuMat& src, GpuMat& dst, int top, int bottom, int left, int right, const Scalar& value = Scalar(), Stream& stream = Stream::Null());
\r
652 //! computes the integral image
\r
653 //! sum will have CV_32S type, but will contain unsigned int values
\r
654 //! supports only CV_8UC1 source type
\r
655 CV_EXPORTS void integral(const GpuMat& src, GpuMat& sum, Stream& stream = Stream::Null());
\r
657 //! buffered version
\r
658 CV_EXPORTS void integralBuffered(const GpuMat& src, GpuMat& sum, GpuMat& buffer, Stream& stream = Stream::Null());
\r
660 //! computes the integral image and integral for the squared image
\r
661 //! sum will have CV_32S type, sqsum - CV32F type
\r
662 //! supports only CV_8UC1 source type
\r
663 CV_EXPORTS void integral(const GpuMat& src, GpuMat& sum, GpuMat& sqsum, Stream& stream = Stream::Null());
\r
665 //! computes squared integral image
\r
666 //! result matrix will have 64F type, but will contain 64U values
\r
667 //! supports source images of 8UC1 type only
\r
668 CV_EXPORTS void sqrIntegral(const GpuMat& src, GpuMat& sqsum, Stream& stream = Stream::Null());
\r
670 //! computes vertical sum, supports only CV_32FC1 images
\r
671 CV_EXPORTS void columnSum(const GpuMat& src, GpuMat& sum);
\r
673 //! computes the standard deviation of integral images
\r
674 //! supports only CV_32SC1 source type and CV_32FC1 sqr type
\r
675 //! output will have CV_32FC1 type
\r
676 CV_EXPORTS void rectStdDev(const GpuMat& src, const GpuMat& sqr, GpuMat& dst, const Rect& rect, Stream& stream = Stream::Null());
\r
678 //! computes Harris cornerness criteria at each image pixel
\r
679 CV_EXPORTS void cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType=BORDER_REFLECT101);
\r
681 //! computes minimum eigen value of 2x2 derivative covariation matrix at each pixel - the cornerness criteria
\r
682 CV_EXPORTS void cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType=BORDER_REFLECT101);
\r
684 //! performs per-element multiplication of two full (not packed) Fourier spectrums
\r
685 //! supports 32FC2 matrixes only (interleaved format)
\r
686 CV_EXPORTS void mulSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c, int flags, bool conjB=false);
\r
688 //! performs per-element multiplication of two full (not packed) Fourier spectrums
\r
689 //! supports 32FC2 matrixes only (interleaved format)
\r
690 CV_EXPORTS void mulAndScaleSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c, int flags,
\r
691 float scale, bool conjB=false);
\r
693 //! Performs a forward or inverse discrete Fourier transform (1D or 2D) of floating point matrix.
\r
694 //! Param dft_size is the size of DFT transform.
\r
696 //! If the source matrix is not continous, then additional copy will be done,
\r
697 //! so to avoid copying ensure the source matrix is continous one. If you want to use
\r
698 //! preallocated output ensure it is continuous too, otherwise it will be reallocated.
\r
700 //! Being implemented via CUFFT real-to-complex transform result contains only non-redundant values
\r
701 //! in CUFFT's format. Result as full complex matrix for such kind of transform cannot be retrieved.
\r
703 //! For complex-to-real transform it is assumed that the source matrix is packed in CUFFT's format.
\r
704 CV_EXPORTS void dft(const GpuMat& src, GpuMat& dst, Size dft_size, int flags=0);
\r
706 //! computes convolution (or cross-correlation) of two images using discrete Fourier transform
\r
707 //! supports source images of 32FC1 type only
\r
708 //! result matrix will have 32FC1 type
\r
709 CV_EXPORTS void convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result,
\r
712 struct CV_EXPORTS ConvolveBuf;
\r
714 //! buffered version
\r
715 CV_EXPORTS void convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result,
\r
716 bool ccorr, ConvolveBuf& buf);
\r
718 struct CV_EXPORTS ConvolveBuf
\r
721 ConvolveBuf(Size image_size, Size templ_size)
\r
722 { create(image_size, templ_size); }
\r
723 void create(Size image_size, Size templ_size);
\r
726 static Size estimateBlockSize(Size result_size, Size templ_size);
\r
727 friend void convolve(const GpuMat&, const GpuMat&, GpuMat&, bool, ConvolveBuf&);
\r
734 GpuMat image_spect, templ_spect, result_spect;
\r
735 GpuMat image_block, templ_block, result_data;
\r
738 //! computes the proximity map for the raster template and the image where the template is searched for
\r
739 CV_EXPORTS void matchTemplate(const GpuMat& image, const GpuMat& templ, GpuMat& result, int method);
\r
741 //! downsamples image
\r
742 CV_EXPORTS void downsample(const GpuMat& src, GpuMat& dst);
\r
744 //! upsamples image
\r
745 CV_EXPORTS void upsample(const GpuMat& src, GpuMat &dst);
\r
747 //! smoothes the source image and downsamples it
\r
748 CV_EXPORTS void pyrDown(const GpuMat& src, GpuMat& dst);
\r
750 //! upsamples the source image and then smoothes it
\r
751 CV_EXPORTS void pyrUp(const GpuMat& src, GpuMat& dst);
\r
753 //! performs linear blending of two images
\r
754 //! to avoid accuracy errors sum of weigths shouldn't be very close to zero
\r
755 CV_EXPORTS void blendLinear(const GpuMat& img1, const GpuMat& img2, const GpuMat& weights1, const GpuMat& weights2,
\r
756 GpuMat& result, Stream& stream = Stream::Null());
\r
758 ////////////////////////////// Matrix reductions //////////////////////////////
\r
760 //! computes mean value and standard deviation of all or selected array elements
\r
761 //! supports only CV_8UC1 type
\r
762 CV_EXPORTS void meanStdDev(const GpuMat& mtx, Scalar& mean, Scalar& stddev);
\r
764 //! computes norm of array
\r
765 //! supports NORM_INF, NORM_L1, NORM_L2
\r
766 //! supports all matrices except 64F
\r
767 CV_EXPORTS double norm(const GpuMat& src1, int normType=NORM_L2);
\r
769 //! computes norm of array
\r
770 //! supports NORM_INF, NORM_L1, NORM_L2
\r
771 //! supports all matrices except 64F
\r
772 CV_EXPORTS double norm(const GpuMat& src1, int normType, GpuMat& buf);
\r
774 //! computes norm of the difference between two arrays
\r
775 //! supports NORM_INF, NORM_L1, NORM_L2
\r
776 //! supports only CV_8UC1 type
\r
777 CV_EXPORTS double norm(const GpuMat& src1, const GpuMat& src2, int normType=NORM_L2);
\r
779 //! computes sum of array elements
\r
780 //! supports only single channel images
\r
781 CV_EXPORTS Scalar sum(const GpuMat& src);
\r
783 //! computes sum of array elements
\r
784 //! supports only single channel images
\r
785 CV_EXPORTS Scalar sum(const GpuMat& src, GpuMat& buf);
\r
787 //! computes sum of array elements absolute values
\r
788 //! supports only single channel images
\r
789 CV_EXPORTS Scalar absSum(const GpuMat& src);
\r
791 //! computes sum of array elements absolute values
\r
792 //! supports only single channel images
\r
793 CV_EXPORTS Scalar absSum(const GpuMat& src, GpuMat& buf);
\r
795 //! computes squared sum of array elements
\r
796 //! supports only single channel images
\r
797 CV_EXPORTS Scalar sqrSum(const GpuMat& src);
\r
799 //! computes squared sum of array elements
\r
800 //! supports only single channel images
\r
801 CV_EXPORTS Scalar sqrSum(const GpuMat& src, GpuMat& buf);
\r
803 //! finds global minimum and maximum array elements and returns their values
\r
804 CV_EXPORTS void minMax(const GpuMat& src, double* minVal, double* maxVal=0, const GpuMat& mask=GpuMat());
\r
806 //! finds global minimum and maximum array elements and returns their values
\r
807 CV_EXPORTS void minMax(const GpuMat& src, double* minVal, double* maxVal, const GpuMat& mask, GpuMat& buf);
\r
809 //! finds global minimum and maximum array elements and returns their values with locations
\r
810 CV_EXPORTS void minMaxLoc(const GpuMat& src, double* minVal, double* maxVal=0, Point* minLoc=0, Point* maxLoc=0,
\r
811 const GpuMat& mask=GpuMat());
\r
813 //! finds global minimum and maximum array elements and returns their values with locations
\r
814 CV_EXPORTS void minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point* minLoc, Point* maxLoc,
\r
815 const GpuMat& mask, GpuMat& valbuf, GpuMat& locbuf);
\r
817 //! counts non-zero array elements
\r
818 CV_EXPORTS int countNonZero(const GpuMat& src);
\r
820 //! counts non-zero array elements
\r
821 CV_EXPORTS int countNonZero(const GpuMat& src, GpuMat& buf);
\r
824 ///////////////////////////// Calibration 3D //////////////////////////////////
\r
826 CV_EXPORTS void transformPoints(const GpuMat& src, const Mat& rvec, const Mat& tvec,
\r
827 GpuMat& dst, Stream& stream = Stream::Null());
\r
829 CV_EXPORTS void projectPoints(const GpuMat& src, const Mat& rvec, const Mat& tvec,
\r
830 const Mat& camera_mat, const Mat& dist_coef, GpuMat& dst,
\r
831 Stream& stream = Stream::Null());
\r
833 CV_EXPORTS void solvePnPRansac(const Mat& object, const Mat& image, const Mat& camera_mat,
\r
834 const Mat& dist_coef, Mat& rvec, Mat& tvec, bool use_extrinsic_guess=false,
\r
835 int num_iters=100, float max_dist=8.0, int min_inlier_count=100,
\r
836 vector<int>* inliers=NULL);
\r
838 //////////////////////////////// Filter Engine ////////////////////////////////
\r
841 The Base Class for 1D or Row-wise Filters
\r
843 This is the base class for linear or non-linear filters that process 1D data.
\r
844 In particular, such filters are used for the "horizontal" filtering parts in separable filters.
\r
846 class CV_EXPORTS BaseRowFilter_GPU
\r
849 BaseRowFilter_GPU(int ksize_, int anchor_) : ksize(ksize_), anchor(anchor_) {}
\r
850 virtual ~BaseRowFilter_GPU() {}
\r
851 virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
\r
856 The Base Class for Column-wise Filters
\r
858 This is the base class for linear or non-linear filters that process columns of 2D arrays.
\r
859 Such filters are used for the "vertical" filtering parts in separable filters.
\r
861 class CV_EXPORTS BaseColumnFilter_GPU
\r
864 BaseColumnFilter_GPU(int ksize_, int anchor_) : ksize(ksize_), anchor(anchor_) {}
\r
865 virtual ~BaseColumnFilter_GPU() {}
\r
866 virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
\r
871 The Base Class for Non-Separable 2D Filters.
\r
873 This is the base class for linear or non-linear 2D filters.
\r
875 class CV_EXPORTS BaseFilter_GPU
\r
878 BaseFilter_GPU(const Size& ksize_, const Point& anchor_) : ksize(ksize_), anchor(anchor_) {}
\r
879 virtual ~BaseFilter_GPU() {}
\r
880 virtual void operator()(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null()) = 0;
\r
886 The Base Class for Filter Engine.
\r
888 The class can be used to apply an arbitrary filtering operation to an image.
\r
889 It contains all the necessary intermediate buffers.
\r
891 class CV_EXPORTS FilterEngine_GPU
\r
894 virtual ~FilterEngine_GPU() {}
\r
896 virtual void apply(const GpuMat& src, GpuMat& dst, Rect roi = Rect(0,0,-1,-1), Stream& stream = Stream::Null()) = 0;
\r
899 //! returns the non-separable filter engine with the specified filter
\r
900 CV_EXPORTS Ptr<FilterEngine_GPU> createFilter2D_GPU(const Ptr<BaseFilter_GPU>& filter2D, int srcType, int dstType);
\r
902 //! returns the separable filter engine with the specified filters
\r
903 CV_EXPORTS Ptr<FilterEngine_GPU> createSeparableFilter_GPU(const Ptr<BaseRowFilter_GPU>& rowFilter,
\r
904 const Ptr<BaseColumnFilter_GPU>& columnFilter, int srcType, int bufType, int dstType);
\r
906 //! returns horizontal 1D box filter
\r
907 //! supports only CV_8UC1 source type and CV_32FC1 sum type
\r
908 CV_EXPORTS Ptr<BaseRowFilter_GPU> getRowSumFilter_GPU(int srcType, int sumType, int ksize, int anchor = -1);
\r
910 //! returns vertical 1D box filter
\r
911 //! supports only CV_8UC1 sum type and CV_32FC1 dst type
\r
912 CV_EXPORTS Ptr<BaseColumnFilter_GPU> getColumnSumFilter_GPU(int sumType, int dstType, int ksize, int anchor = -1);
\r
914 //! returns 2D box filter
\r
915 //! supports CV_8UC1 and CV_8UC4 source type, dst type must be the same as source type
\r
916 CV_EXPORTS Ptr<BaseFilter_GPU> getBoxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1, -1));
\r
918 //! returns box filter engine
\r
919 CV_EXPORTS Ptr<FilterEngine_GPU> createBoxFilter_GPU(int srcType, int dstType, const Size& ksize,
\r
920 const Point& anchor = Point(-1,-1));
\r
922 //! returns 2D morphological filter
\r
923 //! only MORPH_ERODE and MORPH_DILATE are supported
\r
924 //! supports CV_8UC1 and CV_8UC4 types
\r
925 //! kernel must have CV_8UC1 type, one rows and cols == ksize.width * ksize.height
\r
926 CV_EXPORTS Ptr<BaseFilter_GPU> getMorphologyFilter_GPU(int op, int type, const Mat& kernel, const Size& ksize,
\r
927 Point anchor=Point(-1,-1));
\r
929 //! returns morphological filter engine. Only MORPH_ERODE and MORPH_DILATE are supported.
\r
930 CV_EXPORTS Ptr<FilterEngine_GPU> createMorphologyFilter_GPU(int op, int type, const Mat& kernel,
\r
931 const Point& anchor = Point(-1,-1), int iterations = 1);
\r
933 //! returns 2D filter with the specified kernel
\r
934 //! supports CV_8UC1 and CV_8UC4 types
\r
935 CV_EXPORTS Ptr<BaseFilter_GPU> getLinearFilter_GPU(int srcType, int dstType, const Mat& kernel, const Size& ksize,
\r
936 Point anchor = Point(-1, -1));
\r
938 //! returns the non-separable linear filter engine
\r
939 CV_EXPORTS Ptr<FilterEngine_GPU> createLinearFilter_GPU(int srcType, int dstType, const Mat& kernel,
\r
940 const Point& anchor = Point(-1,-1));
\r
942 //! returns the primitive row filter with the specified kernel.
\r
943 //! supports only CV_8UC1, CV_8UC4, CV_16SC1, CV_16SC2, CV_32SC1, CV_32FC1 source type.
\r
944 //! there are two version of algorithm: NPP and OpenCV.
\r
945 //! NPP calls when srcType == CV_8UC1 or srcType == CV_8UC4 and bufType == srcType,
\r
946 //! otherwise calls OpenCV version.
\r
947 //! NPP supports only BORDER_CONSTANT border type.
\r
948 //! OpenCV version supports only CV_32F as buffer depth and
\r
949 //! BORDER_REFLECT101, BORDER_REPLICATE and BORDER_CONSTANT border types.
\r
950 CV_EXPORTS Ptr<BaseRowFilter_GPU> getLinearRowFilter_GPU(int srcType, int bufType, const Mat& rowKernel,
\r
951 int anchor = -1, int borderType = BORDER_CONSTANT);
\r
953 //! returns the primitive column filter with the specified kernel.
\r
954 //! supports only CV_8UC1, CV_8UC4, CV_16SC1, CV_16SC2, CV_32SC1, CV_32FC1 dst type.
\r
955 //! there are two version of algorithm: NPP and OpenCV.
\r
956 //! NPP calls when dstType == CV_8UC1 or dstType == CV_8UC4 and bufType == dstType,
\r
957 //! otherwise calls OpenCV version.
\r
958 //! NPP supports only BORDER_CONSTANT border type.
\r
959 //! OpenCV version supports only CV_32F as buffer depth and
\r
960 //! BORDER_REFLECT101, BORDER_REPLICATE and BORDER_CONSTANT border types.
\r
961 CV_EXPORTS Ptr<BaseColumnFilter_GPU> getLinearColumnFilter_GPU(int bufType, int dstType, const Mat& columnKernel,
\r
962 int anchor = -1, int borderType = BORDER_CONSTANT);
\r
964 //! returns the separable linear filter engine
\r
965 CV_EXPORTS Ptr<FilterEngine_GPU> createSeparableLinearFilter_GPU(int srcType, int dstType, const Mat& rowKernel,
\r
966 const Mat& columnKernel, const Point& anchor = Point(-1,-1), int rowBorderType = BORDER_DEFAULT,
\r
967 int columnBorderType = -1);
\r
969 //! returns filter engine for the generalized Sobel operator
\r
970 CV_EXPORTS Ptr<FilterEngine_GPU> createDerivFilter_GPU(int srcType, int dstType, int dx, int dy, int ksize,
\r
971 int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
\r
973 //! returns the Gaussian filter engine
\r
974 CV_EXPORTS Ptr<FilterEngine_GPU> createGaussianFilter_GPU(int type, Size ksize, double sigma1, double sigma2 = 0,
\r
975 int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1);
\r
977 //! returns maximum filter
\r
978 CV_EXPORTS Ptr<BaseFilter_GPU> getMaxFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1,-1));
\r
980 //! returns minimum filter
\r
981 CV_EXPORTS Ptr<BaseFilter_GPU> getMinFilter_GPU(int srcType, int dstType, const Size& ksize, Point anchor = Point(-1,-1));
\r
983 //! smooths the image using the normalized box filter
\r
984 //! supports CV_8UC1, CV_8UC4 types
\r
985 CV_EXPORTS void boxFilter(const GpuMat& src, GpuMat& dst, int ddepth, Size ksize, Point anchor = Point(-1,-1), Stream& stream = Stream::Null());
\r
987 //! a synonym for normalized box filter
\r
988 static inline void blur(const GpuMat& src, GpuMat& dst, Size ksize, Point anchor = Point(-1,-1), Stream& stream = Stream::Null()) { boxFilter(src, dst, -1, ksize, anchor, stream); }
\r
990 //! erodes the image (applies the local minimum operator)
\r
991 CV_EXPORTS void erode( const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor = Point(-1, -1), int iterations = 1, Stream& stream = Stream::Null());
\r
993 //! dilates the image (applies the local maximum operator)
\r
994 CV_EXPORTS void dilate( const GpuMat& src, GpuMat& dst, const Mat& kernel, Point anchor = Point(-1, -1), int iterations = 1, Stream& stream = Stream::Null());
\r
996 //! applies an advanced morphological operation to the image
\r
997 CV_EXPORTS void morphologyEx( const GpuMat& src, GpuMat& dst, int op, const Mat& kernel, Point anchor = Point(-1, -1), int iterations = 1, Stream& stream = Stream::Null());
\r
999 //! applies non-separable 2D linear filter to the image
\r
1000 CV_EXPORTS void filter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernel, Point anchor=Point(-1,-1), Stream& stream = Stream::Null());
\r
1002 //! applies separable 2D linear filter to the image
\r
1003 CV_EXPORTS void sepFilter2D(const GpuMat& src, GpuMat& dst, int ddepth, const Mat& kernelX, const Mat& kernelY,
\r
1004 Point anchor = Point(-1,-1), int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1, Stream& stream = Stream::Null());
\r
1006 //! applies generalized Sobel operator to the image
\r
1007 CV_EXPORTS void Sobel(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, int ksize = 3, double scale = 1,
\r
1008 int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1, Stream& stream = Stream::Null());
\r
1010 //! applies the vertical or horizontal Scharr operator to the image
\r
1011 CV_EXPORTS void Scharr(const GpuMat& src, GpuMat& dst, int ddepth, int dx, int dy, double scale = 1,
\r
1012 int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1, Stream& stream = Stream::Null());
\r
1014 //! smooths the image using Gaussian filter.
\r
1015 CV_EXPORTS void GaussianBlur(const GpuMat& src, GpuMat& dst, Size ksize, double sigma1, double sigma2 = 0,
\r
1016 int rowBorderType = BORDER_DEFAULT, int columnBorderType = -1, Stream& stream = Stream::Null());
\r
1018 //! applies Laplacian operator to the image
\r
1019 //! supports only ksize = 1 and ksize = 3
\r
1020 CV_EXPORTS void Laplacian(const GpuMat& src, GpuMat& dst, int ddepth, int ksize = 1, double scale = 1, Stream& stream = Stream::Null());
\r
1022 //////////////////////////////// Image Labeling ////////////////////////////////
\r
1024 //!performs labeling via graph cuts
\r
1025 CV_EXPORTS void graphcut(GpuMat& terminals, GpuMat& leftTransp, GpuMat& rightTransp, GpuMat& top, GpuMat& bottom, GpuMat& labels, GpuMat& buf, Stream& stream = Stream::Null());
\r
1027 ////////////////////////////////// Histograms //////////////////////////////////
\r
1029 //! Compute levels with even distribution. levels will have 1 row and nLevels cols and CV_32SC1 type.
\r
1030 CV_EXPORTS void evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperLevel);
\r
1031 //! Calculates histogram with evenly distributed bins for signle channel source.
\r
1032 //! Supports CV_8UC1, CV_16UC1 and CV_16SC1 source types.
\r
1033 //! Output hist will have one row and histSize cols and CV_32SC1 type.
\r
1034 CV_EXPORTS void histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null());
\r
1035 //! Calculates histogram with evenly distributed bins for four-channel source.
\r
1036 //! All channels of source are processed separately.
\r
1037 //! Supports CV_8UC4, CV_16UC4 and CV_16SC4 source types.
\r
1038 //! Output hist[i] will have one row and histSize[i] cols and CV_32SC1 type.
\r
1039 CV_EXPORTS void histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null());
\r
1040 //! Calculates histogram with bins determined by levels array.
\r
1041 //! levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.
\r
1042 //! Supports CV_8UC1, CV_16UC1, CV_16SC1 and CV_32FC1 source types.
\r
1043 //! Output hist will have one row and (levels.cols-1) cols and CV_32SC1 type.
\r
1044 CV_EXPORTS void histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream = Stream::Null());
\r
1045 //! Calculates histogram with bins determined by levels array.
\r
1046 //! All levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.
\r
1047 //! All channels of source are processed separately.
\r
1048 //! Supports CV_8UC4, CV_16UC4, CV_16SC4 and CV_32FC4 source types.
\r
1049 //! Output hist[i] will have one row and (levels[i].cols-1) cols and CV_32SC1 type.
\r
1050 CV_EXPORTS void histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], Stream& stream = Stream::Null());
\r
1052 //////////////////////////////// StereoBM_GPU ////////////////////////////////
\r
1054 class CV_EXPORTS StereoBM_GPU
\r
1057 enum { BASIC_PRESET = 0, PREFILTER_XSOBEL = 1 };
\r
1059 enum { DEFAULT_NDISP = 64, DEFAULT_WINSZ = 19 };
\r
1061 //! the default constructor
\r
1063 //! the full constructor taking the camera-specific preset, number of disparities and the SAD window size. ndisparities must be multiple of 8.
\r
1064 StereoBM_GPU(int preset, int ndisparities = DEFAULT_NDISP, int winSize = DEFAULT_WINSZ);
\r
1066 //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair
\r
1067 //! Output disparity has CV_8U type.
\r
1068 void operator() ( const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null());
\r
1070 //! Some heuristics that tries to estmate
\r
1071 // if current GPU will be faster than CPU in this algorithm.
\r
1072 // It queries current active device.
\r
1073 static bool checkIfGpuCallReasonable();
\r
1079 // If avergeTexThreshold == 0 => post procesing is disabled
\r
1080 // If avergeTexThreshold != 0 then disparity is set 0 in each point (x,y) where for left image
\r
1081 // SumOfHorizontalGradiensInWindow(x, y, winSize) < (winSize * winSize) * avergeTexThreshold
\r
1082 // i.e. input left image is low textured.
\r
1083 float avergeTexThreshold;
\r
1085 GpuMat minSSD, leBuf, riBuf;
\r
1088 ////////////////////////// StereoBeliefPropagation ///////////////////////////
\r
1089 // "Efficient Belief Propagation for Early Vision"
\r
1092 class CV_EXPORTS StereoBeliefPropagation
\r
1095 enum { DEFAULT_NDISP = 64 };
\r
1096 enum { DEFAULT_ITERS = 5 };
\r
1097 enum { DEFAULT_LEVELS = 5 };
\r
1099 static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels);
\r
1101 //! the default constructor
\r
1102 explicit StereoBeliefPropagation(int ndisp = DEFAULT_NDISP,
\r
1103 int iters = DEFAULT_ITERS,
\r
1104 int levels = DEFAULT_LEVELS,
\r
1105 int msg_type = CV_32F);
\r
1107 //! the full constructor taking the number of disparities, number of BP iterations on each level,
\r
1108 //! number of levels, truncation of data cost, data weight,
\r
1109 //! truncation of discontinuity cost and discontinuity single jump
\r
1110 //! DataTerm = data_weight * min(fabs(I2-I1), max_data_term)
\r
1111 //! DiscTerm = min(disc_single_jump * fabs(f1-f2), max_disc_term)
\r
1112 //! please see paper for more details
\r
1113 StereoBeliefPropagation(int ndisp, int iters, int levels,
\r
1114 float max_data_term, float data_weight,
\r
1115 float max_disc_term, float disc_single_jump,
\r
1116 int msg_type = CV_32F);
\r
1118 //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair,
\r
1119 //! if disparity is empty output type will be CV_16S else output type will be disparity.type().
\r
1120 void operator()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null());
\r
1123 //! version for user specified data term
\r
1124 void operator()(const GpuMat& data, GpuMat& disparity, Stream& stream = Stream::Null());
\r
1131 float max_data_term;
\r
1132 float data_weight;
\r
1133 float max_disc_term;
\r
1134 float disc_single_jump;
\r
1138 GpuMat u, d, l, r, u2, d2, l2, r2;
\r
1139 std::vector<GpuMat> datas;
\r
1143 /////////////////////////// StereoConstantSpaceBP ///////////////////////////
\r
1144 // "A Constant-Space Belief Propagation Algorithm for Stereo Matching"
\r
1145 // Qingxiong Yang, Liang Wang, Narendra Ahuja
\r
1146 // http://vision.ai.uiuc.edu/~qyang6/
\r
1148 class CV_EXPORTS StereoConstantSpaceBP
\r
1151 enum { DEFAULT_NDISP = 128 };
\r
1152 enum { DEFAULT_ITERS = 8 };
\r
1153 enum { DEFAULT_LEVELS = 4 };
\r
1154 enum { DEFAULT_NR_PLANE = 4 };
\r
1156 static void estimateRecommendedParams(int width, int height, int& ndisp, int& iters, int& levels, int& nr_plane);
\r
1158 //! the default constructor
\r
1159 explicit StereoConstantSpaceBP(int ndisp = DEFAULT_NDISP,
\r
1160 int iters = DEFAULT_ITERS,
\r
1161 int levels = DEFAULT_LEVELS,
\r
1162 int nr_plane = DEFAULT_NR_PLANE,
\r
1163 int msg_type = CV_32F);
\r
1165 //! the full constructor taking the number of disparities, number of BP iterations on each level,
\r
1166 //! number of levels, number of active disparity on the first level, truncation of data cost, data weight,
\r
1167 //! truncation of discontinuity cost, discontinuity single jump and minimum disparity threshold
\r
1168 StereoConstantSpaceBP(int ndisp, int iters, int levels, int nr_plane,
\r
1169 float max_data_term, float data_weight, float max_disc_term, float disc_single_jump,
\r
1170 int min_disp_th = 0,
\r
1171 int msg_type = CV_32F);
\r
1173 //! the stereo correspondence operator. Finds the disparity for the specified rectified stereo pair,
\r
1174 //! if disparity is empty output type will be CV_16S else output type will be disparity.type().
\r
1175 void operator()(const GpuMat& left, const GpuMat& right, GpuMat& disparity, Stream& stream = Stream::Null());
\r
1184 float max_data_term;
\r
1185 float data_weight;
\r
1186 float max_disc_term;
\r
1187 float disc_single_jump;
\r
1193 bool use_local_init_data_cost;
\r
1195 GpuMat u[2], d[2], l[2], r[2];
\r
1196 GpuMat disp_selected_pyr[2];
\r
1199 GpuMat data_cost_selected;
\r
1206 /////////////////////////// DisparityBilateralFilter ///////////////////////////
\r
1207 // Disparity map refinement using joint bilateral filtering given a single color image.
\r
1208 // Qingxiong Yang, Liang Wang, Narendra Ahuja
\r
1209 // http://vision.ai.uiuc.edu/~qyang6/
\r
1211 class CV_EXPORTS DisparityBilateralFilter
\r
1214 enum { DEFAULT_NDISP = 64 };
\r
1215 enum { DEFAULT_RADIUS = 3 };
\r
1216 enum { DEFAULT_ITERS = 1 };
\r
1218 //! the default constructor
\r
1219 explicit DisparityBilateralFilter(int ndisp = DEFAULT_NDISP, int radius = DEFAULT_RADIUS, int iters = DEFAULT_ITERS);
\r
1221 //! the full constructor taking the number of disparities, filter radius,
\r
1222 //! number of iterations, truncation of data continuity, truncation of disparity continuity
\r
1223 //! and filter range sigma
\r
1224 DisparityBilateralFilter(int ndisp, int radius, int iters, float edge_threshold, float max_disc_threshold, float sigma_range);
\r
1226 //! the disparity map refinement operator. Refine disparity map using joint bilateral filtering given a single color image.
\r
1227 //! disparity must have CV_8U or CV_16S type, image must have CV_8UC1 or CV_8UC3 type.
\r
1228 void operator()(const GpuMat& disparity, const GpuMat& image, GpuMat& dst, Stream& stream = Stream::Null());
\r
1235 float edge_threshold;
\r
1236 float max_disc_threshold;
\r
1237 float sigma_range;
\r
1239 GpuMat table_color;
\r
1240 GpuMat table_space;
\r
1244 //////////////// HOG (Histogram-of-Oriented-Gradients) Descriptor and Object Detector //////////////
\r
1246 struct CV_EXPORTS HOGDescriptor
\r
1248 enum { DEFAULT_WIN_SIGMA = -1 };
\r
1249 enum { DEFAULT_NLEVELS = 64 };
\r
1250 enum { DESCR_FORMAT_ROW_BY_ROW, DESCR_FORMAT_COL_BY_COL };
\r
1252 HOGDescriptor(Size win_size=Size(64, 128), Size block_size=Size(16, 16),
\r
1253 Size block_stride=Size(8, 8), Size cell_size=Size(8, 8),
\r
1254 int nbins=9, double win_sigma=DEFAULT_WIN_SIGMA,
\r
1255 double threshold_L2hys=0.2, bool gamma_correction=true,
\r
1256 int nlevels=DEFAULT_NLEVELS);
\r
1258 size_t getDescriptorSize() const;
\r
1259 size_t getBlockHistogramSize() const;
\r
1261 void setSVMDetector(const vector<float>& detector);
\r
1263 static vector<float> getDefaultPeopleDetector();
\r
1264 static vector<float> getPeopleDetector48x96();
\r
1265 static vector<float> getPeopleDetector64x128();
\r
1267 void detect(const GpuMat& img, vector<Point>& found_locations,
\r
1268 double hit_threshold=0, Size win_stride=Size(),
\r
1269 Size padding=Size());
\r
1271 void detectMultiScale(const GpuMat& img, vector<Rect>& found_locations,
\r
1272 double hit_threshold=0, Size win_stride=Size(),
\r
1273 Size padding=Size(), double scale0=1.05,
\r
1274 int group_threshold=2);
\r
1276 void getDescriptors(const GpuMat& img, Size win_stride,
\r
1277 GpuMat& descriptors,
\r
1278 int descr_format=DESCR_FORMAT_COL_BY_COL);
\r
1282 Size block_stride;
\r
1286 double threshold_L2hys;
\r
1287 bool gamma_correction;
\r
1291 void computeBlockHistograms(const GpuMat& img);
\r
1292 void computeGradient(const GpuMat& img, GpuMat& grad, GpuMat& qangle);
\r
1294 double getWinSigma() const;
\r
1295 bool checkDetectorSize() const;
\r
1297 static int numPartsWithin(int size, int part_size, int stride);
\r
1298 static Size numPartsWithin(Size size, Size part_size, Size stride);
\r
1300 // Coefficients of the separating plane
\r
1304 // Results of the last classification step
\r
1305 GpuMat labels, labels_buf;
\r
1308 // Results of the last histogram evaluation step
\r
1309 GpuMat block_hists, block_hists_buf;
\r
1311 // Gradients conputation results
\r
1312 GpuMat grad, qangle, grad_buf, qangle_buf;
\r
1314 // returns subbuffer with required size, reallocates buffer if nessesary.
\r
1315 static GpuMat getBuffer(const Size& sz, int type, GpuMat& buf);
\r
1316 static GpuMat getBuffer(int rows, int cols, int type, GpuMat& buf);
\r
1318 std::vector<GpuMat> image_scales;
\r
1322 ////////////////////////////////// BruteForceMatcher //////////////////////////////////
\r
1324 class CV_EXPORTS BruteForceMatcher_GPU_base
\r
1327 enum DistType {L1Dist = 0, L2Dist, HammingDist};
\r
1329 explicit BruteForceMatcher_GPU_base(DistType distType = L2Dist);
\r
1331 // Add descriptors to train descriptor collection.
\r
1332 void add(const std::vector<GpuMat>& descCollection);
\r
1334 // Get train descriptors collection.
\r
1335 const std::vector<GpuMat>& getTrainDescriptors() const;
\r
1337 // Clear train descriptors collection.
\r
1340 // Return true if there are not train descriptors in collection.
\r
1341 bool empty() const;
\r
1343 // Return true if the matcher supports mask in match methods.
\r
1344 bool isMaskSupported() const;
\r
1346 // Find one best match for each query descriptor.
\r
1347 // trainIdx.at<int>(0, queryIdx) will contain best train index for queryIdx
\r
1348 // distance.at<float>(0, queryIdx) will contain distance
\r
1349 void matchSingle(const GpuMat& queryDescs, const GpuMat& trainDescs,
\r
1350 GpuMat& trainIdx, GpuMat& distance,
\r
1351 const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
\r
1353 // Download trainIdx and distance to CPU vector with DMatch
\r
1354 static void matchDownload(const GpuMat& trainIdx, const GpuMat& distance, std::vector<DMatch>& matches);
\r
1356 // Find one best match for each query descriptor.
\r
1357 void match(const GpuMat& queryDescs, const GpuMat& trainDescs, std::vector<DMatch>& matches,
\r
1358 const GpuMat& mask = GpuMat());
\r
1360 // Make gpu collection of trains and masks in suitable format for matchCollection function
\r
1361 void makeGpuCollection(GpuMat& trainCollection, GpuMat& maskCollection,
\r
1362 const vector<GpuMat>& masks = std::vector<GpuMat>());
\r
1364 // Find one best match from train collection for each query descriptor.
\r
1365 // trainIdx.at<int>(0, queryIdx) will contain best train index for queryIdx
\r
1366 // imgIdx.at<int>(0, queryIdx) will contain best image index for queryIdx
\r
1367 // distance.at<float>(0, queryIdx) will contain distance
\r
1368 void matchCollection(const GpuMat& queryDescs, const GpuMat& trainCollection,
\r
1369 GpuMat& trainIdx, GpuMat& imgIdx, GpuMat& distance,
\r
1370 const GpuMat& maskCollection, Stream& stream = Stream::Null());
\r
1372 // Download trainIdx, imgIdx and distance to CPU vector with DMatch
\r
1373 static void matchDownload(const GpuMat& trainIdx, const GpuMat& imgIdx, const GpuMat& distance,
\r
1374 std::vector<DMatch>& matches);
\r
1376 // Find one best match from train collection for each query descriptor.
\r
1377 void match(const GpuMat& queryDescs, std::vector<DMatch>& matches,
\r
1378 const std::vector<GpuMat>& masks = std::vector<GpuMat>());
\r
1380 // Find k best matches for each query descriptor (in increasing order of distances).
\r
1381 // trainIdx.at<int>(queryIdx, i) will contain index of i'th best trains (i < k).
\r
1382 // distance.at<float>(queryIdx, i) will contain distance.
\r
1383 // allDist is a buffer to store all distance between query descriptors and train descriptors
\r
1384 // it have size (nQuery,nTrain) and CV_32F type
\r
1385 // allDist.at<float>(queryIdx, trainIdx) will contain FLT_MAX, if trainIdx is one from k best,
\r
1386 // otherwise it will contain distance between queryIdx and trainIdx descriptors
\r
1387 void knnMatch(const GpuMat& queryDescs, const GpuMat& trainDescs,
\r
1388 GpuMat& trainIdx, GpuMat& distance, GpuMat& allDist, int k, const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
\r
1390 // Download trainIdx and distance to CPU vector with DMatch
\r
1391 // compactResult is used when mask is not empty. If compactResult is false matches
\r
1392 // vector will have the same size as queryDescriptors rows. If compactResult is true
\r
1393 // matches vector will not contain matches for fully masked out query descriptors.
\r
1394 static void knnMatchDownload(const GpuMat& trainIdx, const GpuMat& distance,
\r
1395 std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
\r
1397 // Find k best matches for each query descriptor (in increasing order of distances).
\r
1398 // compactResult is used when mask is not empty. If compactResult is false matches
\r
1399 // vector will have the same size as queryDescriptors rows. If compactResult is true
\r
1400 // matches vector will not contain matches for fully masked out query descriptors.
\r
1401 void knnMatch(const GpuMat& queryDescs, const GpuMat& trainDescs,
\r
1402 std::vector< std::vector<DMatch> >& matches, int k, const GpuMat& mask = GpuMat(),
\r
1403 bool compactResult = false);
\r
1405 // Find k best matches for each query descriptor (in increasing order of distances).
\r
1406 // compactResult is used when mask is not empty. If compactResult is false matches
\r
1407 // vector will have the same size as queryDescriptors rows. If compactResult is true
\r
1408 // matches vector will not contain matches for fully masked out query descriptors.
\r
1409 void knnMatch(const GpuMat& queryDescs, std::vector< std::vector<DMatch> >& matches, int knn,
\r
1410 const std::vector<GpuMat>& masks = std::vector<GpuMat>(), bool compactResult = false );
\r
1412 // Find best matches for each query descriptor which have distance less than maxDistance.
\r
1413 // nMatches.at<unsigned int>(0, queruIdx) will contain matches count for queryIdx.
\r
1414 // carefully nMatches can be greater than trainIdx.cols - it means that matcher didn't find all matches,
\r
1415 // because it didn't have enough memory.
\r
1416 // trainIdx.at<int>(queruIdx, i) will contain ith train index (i < min(nMatches.at<unsigned int>(0, queruIdx), trainIdx.cols))
\r
1417 // distance.at<int>(queruIdx, i) will contain ith distance (i < min(nMatches.at<unsigned int>(0, queruIdx), trainIdx.cols))
\r
1418 // If trainIdx is empty, then trainIdx and distance will be created with size nQuery x nTrain,
\r
1419 // otherwize user can pass own allocated trainIdx and distance with size nQuery x nMaxMatches
\r
1420 // Matches doesn't sorted.
\r
1421 void radiusMatch(const GpuMat& queryDescs, const GpuMat& trainDescs,
\r
1422 GpuMat& trainIdx, GpuMat& nMatches, GpuMat& distance, float maxDistance,
\r
1423 const GpuMat& mask = GpuMat(), Stream& stream = Stream::Null());
\r
1425 // Download trainIdx, nMatches and distance to CPU vector with DMatch.
\r
1426 // matches will be sorted in increasing order of distances.
\r
1427 // compactResult is used when mask is not empty. If compactResult is false matches
\r
1428 // vector will have the same size as queryDescriptors rows. If compactResult is true
\r
1429 // matches vector will not contain matches for fully masked out query descriptors.
\r
1430 static void radiusMatchDownload(const GpuMat& trainIdx, const GpuMat& nMatches, const GpuMat& distance,
\r
1431 std::vector< std::vector<DMatch> >& matches, bool compactResult = false);
\r
1433 // Find best matches for each query descriptor which have distance less than maxDistance
\r
1434 // in increasing order of distances).
\r
1435 void radiusMatch(const GpuMat& queryDescs, const GpuMat& trainDescs,
\r
1436 std::vector< std::vector<DMatch> >& matches, float maxDistance,
\r
1437 const GpuMat& mask = GpuMat(), bool compactResult = false);
\r
1439 // Find best matches from train collection for each query descriptor which have distance less than
\r
1440 // maxDistance (in increasing order of distances).
\r
1441 void radiusMatch(const GpuMat& queryDescs, std::vector< std::vector<DMatch> >& matches, float maxDistance,
\r
1442 const std::vector<GpuMat>& masks = std::vector<GpuMat>(), bool compactResult = false);
\r
1444 DistType distType;
\r
1447 std::vector<GpuMat> trainDescCollection;
\r
1450 template <class Distance>
\r
1451 class CV_EXPORTS BruteForceMatcher_GPU;
\r
1453 template <typename T>
\r
1454 class CV_EXPORTS BruteForceMatcher_GPU< L1<T> > : public BruteForceMatcher_GPU_base
\r
1457 explicit BruteForceMatcher_GPU() : BruteForceMatcher_GPU_base(L1Dist) {}
\r
1458 explicit BruteForceMatcher_GPU(L1<T> /*d*/) : BruteForceMatcher_GPU_base(L1Dist) {}
\r
1460 template <typename T>
\r
1461 class CV_EXPORTS BruteForceMatcher_GPU< L2<T> > : public BruteForceMatcher_GPU_base
\r
1464 explicit BruteForceMatcher_GPU() : BruteForceMatcher_GPU_base(L2Dist) {}
\r
1465 explicit BruteForceMatcher_GPU(L2<T> /*d*/) : BruteForceMatcher_GPU_base(L2Dist) {}
\r
1467 template <> class CV_EXPORTS BruteForceMatcher_GPU< HammingLUT > : public BruteForceMatcher_GPU_base
\r
1470 explicit BruteForceMatcher_GPU() : BruteForceMatcher_GPU_base(HammingDist) {}
\r
1471 explicit BruteForceMatcher_GPU(HammingLUT /*d*/) : BruteForceMatcher_GPU_base(HammingDist) {}
\r
1473 template <> class CV_EXPORTS BruteForceMatcher_GPU< Hamming > : public BruteForceMatcher_GPU_base
\r
1476 explicit BruteForceMatcher_GPU() : BruteForceMatcher_GPU_base(HammingDist) {}
\r
1477 explicit BruteForceMatcher_GPU(Hamming /*d*/) : BruteForceMatcher_GPU_base(HammingDist) {}
\r
1480 ////////////////////////////////// CascadeClassifier_GPU //////////////////////////////////////////
\r
1481 // The cascade classifier class for object detection.
\r
1482 class CV_EXPORTS CascadeClassifier_GPU
\r
1485 CascadeClassifier_GPU();
\r
1486 CascadeClassifier_GPU(const string& filename);
\r
1487 ~CascadeClassifier_GPU();
\r
1489 bool empty() const;
\r
1490 bool load(const string& filename);
\r
1493 /* returns number of detected objects */
\r
1494 int detectMultiScale( const GpuMat& image, GpuMat& objectsBuf, double scaleFactor=1.2, int minNeighbors=4, Size minSize=Size());
\r
1496 bool findLargestObject;
\r
1497 bool visualizeInPlace;
\r
1499 Size getClassifierSize() const;
\r
1502 struct CascadeClassifierImpl;
\r
1503 CascadeClassifierImpl* impl;
\r
1506 ////////////////////////////////// SURF //////////////////////////////////////////
\r
1508 class CV_EXPORTS SURF_GPU : public CvSURFParams
\r
1511 enum KeypointLayout
\r
1522 //! the default constructor
\r
1524 //! the full constructor taking all the necessary parameters
\r
1525 explicit SURF_GPU(double _hessianThreshold, int _nOctaves=4,
\r
1526 int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f, bool _upright = false);
\r
1528 //! returns the descriptor size in float's (64 or 128)
\r
1529 int descriptorSize() const;
\r
1531 //! upload host keypoints to device memory
\r
1532 void uploadKeypoints(const vector<KeyPoint>& keypoints, GpuMat& keypointsGPU);
\r
1533 //! download keypoints from device to host memory
\r
1534 void downloadKeypoints(const GpuMat& keypointsGPU, vector<KeyPoint>& keypoints);
\r
1536 //! download descriptors from device to host memory
\r
1537 void downloadDescriptors(const GpuMat& descriptorsGPU, vector<float>& descriptors);
\r
1539 //! finds the keypoints using fast hessian detector used in SURF
\r
1540 //! supports CV_8UC1 images
\r
1541 //! keypoints will have nFeature cols and 6 rows
\r
1542 //! keypoints.ptr<float>(SF_X)[i] will contain x coordinate of i'th feature
\r
1543 //! keypoints.ptr<float>(SF_Y)[i] will contain y coordinate of i'th feature
\r
1544 //! keypoints.ptr<float>(SF_LAPLACIAN)[i] will contain laplacian sign of i'th feature
\r
1545 //! keypoints.ptr<float>(SF_SIZE)[i] will contain size of i'th feature
\r
1546 //! keypoints.ptr<float>(SF_DIR)[i] will contain orientation of i'th feature
\r
1547 //! keypoints.ptr<float>(SF_HESSIAN)[i] will contain response of i'th feature
\r
1548 void operator()(const GpuMat& img, const GpuMat& mask, GpuMat& keypoints);
\r
1549 //! finds the keypoints and computes their descriptors.
\r
1550 //! Optionally it can compute descriptors for the user-provided keypoints and recompute keypoints direction
\r
1551 void operator()(const GpuMat& img, const GpuMat& mask, GpuMat& keypoints, GpuMat& descriptors,
\r
1552 bool useProvidedKeypoints = false);
\r
1554 void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
\r
1555 void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, GpuMat& descriptors,
\r
1556 bool useProvidedKeypoints = false);
\r
1558 void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, std::vector<float>& descriptors,
\r
1559 bool useProvidedKeypoints = false);
\r
1561 //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
\r
1562 float keypointsRatio;
\r
1564 GpuMat sum, mask1, maskSum, intBuffer;
\r
1566 GpuMat det, trace;
\r
1568 GpuMat maxPosBuffer;
\r
1573 //! Speckle filtering - filters small connected components on diparity image.
\r
1574 //! It sets pixel (x,y) to newVal if it coresponds to small CC with size < maxSpeckleSize.
\r
1575 //! Threshold for border between CC is diffThreshold;
\r
1576 CV_EXPORTS void filterSpeckles( Mat& img, uchar newVal, int maxSpeckleSize, uchar diffThreshold, Mat& buf);
\r
1579 #include "opencv2/gpu/matrix_operations.hpp"
\r
1581 #endif /* __OPENCV_GPU_HPP__ */
\r