#include "precomp.hpp"
-#ifdef HAVE_EIGEN2
-#include <Eigen/Array>
-#endif
-
-//#define _KDTREE
-
using namespace std;
-namespace cv
-{
-void drawMatches( const Mat& img1, const vector<KeyPoint>& keypoints1,
- const Mat& img2,const vector<KeyPoint>& keypoints2,
- const vector<int>& matches, Mat& outImg,
- const Scalar& matchColor, const Scalar& singlePointColor,
- const vector<char>& matchesMask, int flags )
+namespace cv
{
- Size size( img1.cols + img2.cols, MAX(img1.rows, img2.rows) );
- if( flags & DrawMatchesFlags::DRAW_OVER_OUTIMG )
- {
- if( size.width > outImg.cols || size.height > outImg.rows )
- CV_Error( CV_StsBadSize, "outImg has size less than need to draw img1 and img2 together" );
- }
- else
- {
- outImg.create( size, CV_MAKETYPE(img1.depth(), 3) );
- Mat outImg1 = outImg( Rect(0, 0, img1.cols, img1.rows) );
- cvtColor( img1, outImg1, CV_GRAY2RGB );
- Mat outImg2 = outImg( Rect(img1.cols, 0, img2.cols, img2.rows) );
- cvtColor( img2, outImg2, CV_GRAY2RGB );
- }
-
- RNG rng;
- // draw keypoints
- if( !(flags & DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS) )
- {
- bool isRandSinglePointColor = singlePointColor == Scalar::all(-1);
- for( vector<KeyPoint>::const_iterator it = keypoints1.begin(); it < keypoints1.end(); ++it )
- {
- circle( outImg, it->pt, 3, isRandSinglePointColor ?
- Scalar(rng.uniform(0, 256), rng.uniform(0, 256), rng.uniform(0, 256)) : singlePointColor );
- }
- for( vector<KeyPoint>::const_iterator it = keypoints2.begin(); it < keypoints2.end(); ++it )
- {
- Point p = it->pt;
- circle( outImg, Point(p.x+img1.cols, p.y), 3, isRandSinglePointColor ?
- Scalar(rng.uniform(0, 256), rng.uniform(0, 256), rng.uniform(0, 256)) : singlePointColor );
- }
- }
-
- // draw matches
- bool isRandMatchColor = matchColor == Scalar::all(-1);
- if( matches.size() != keypoints1.size() )
- CV_Error( CV_StsBadSize, "matches must have the same size as keypoints1" );
- if( !matchesMask.empty() && matchesMask.size() != keypoints1.size() )
- CV_Error( CV_StsBadSize, "mask must have the same size as keypoints1" );
- vector<int>::const_iterator mit = matches.begin();
- for( int i1 = 0; mit != matches.end(); ++mit, i1++ )
- {
- if( (matchesMask.empty() || matchesMask[i1] ) && *mit >= 0 )
- {
- Point2f pt1 = keypoints1[i1].pt,
- pt2 = keypoints2[*mit].pt,
- dpt2 = Point2f( std::min(pt2.x+img1.cols, float(outImg.cols-1)), pt2.y );
- Scalar randColor( rng.uniform(0, 256), rng.uniform(0, 256), rng.uniform(0, 256) );
- circle( outImg, pt1, 3, isRandMatchColor ? randColor : matchColor );
- circle( outImg, dpt2, 3, isRandMatchColor ? randColor : matchColor );
- line( outImg, pt1, dpt2, isRandMatchColor ? randColor : matchColor );
- }
- }
-}
/****************************************************************************************\
* DescriptorExtractor *
/*
* DescriptorExtractor
*/
-struct RoiPredicate
-{
- RoiPredicate(float _minX, float _minY, float _maxX, float _maxY)
- : minX(_minX), minY(_minY), maxX(_maxX), maxY(_maxY)
- {}
-
- bool operator()( const KeyPoint& keyPt) const
- {
- Point2f pt = keyPt.pt;
- return (pt.x < minX) || (pt.x >= maxX) || (pt.y < minY) || (pt.y >= maxY);
- }
-
- float minX, minY, maxX, maxY;
-};
-
-void DescriptorExtractor::removeBorderKeypoints( vector<KeyPoint>& keypoints,
- Size imageSize, int borderPixels )
-{
- keypoints.erase( remove_if(keypoints.begin(), keypoints.end(),
- RoiPredicate((float)borderPixels, (float)borderPixels,
- (float)(imageSize.width - borderPixels),
- (float)(imageSize.height - borderPixels))),
- keypoints.end());
-}
-
-/****************************************************************************************\
-* SiftDescriptorExtractor *
-\****************************************************************************************/
-SiftDescriptorExtractor::SiftDescriptorExtractor( double magnification, bool isNormalize, bool recalculateAngles,
- int nOctaves, int nOctaveLayers, int firstOctave, int angleMode )
- : sift( magnification, isNormalize, recalculateAngles, nOctaves, nOctaveLayers, firstOctave, angleMode )
+DescriptorExtractor::~DescriptorExtractor()
{}
-void SiftDescriptorExtractor::compute( const Mat& image,
- vector<KeyPoint>& keypoints,
- Mat& descriptors) const
-{
- bool useProvidedKeypoints = true;
- sift(image, Mat(), keypoints, descriptors, useProvidedKeypoints);
-}
-
-void SiftDescriptorExtractor::read (const FileNode &fn)
+void DescriptorExtractor::compute( const Mat& image, vector<KeyPoint>& keypoints, Mat& descriptors ) const
{
- double magnification = fn["magnification"];
- bool isNormalize = (int)fn["isNormalize"] != 0;
- bool recalculateAngles = (int)fn["recalculateAngles"] != 0;
- int nOctaves = fn["nOctaves"];
- int nOctaveLayers = fn["nOctaveLayers"];
- int firstOctave = fn["firstOctave"];
- int angleMode = fn["angleMode"];
-
- sift = SIFT( magnification, isNormalize, recalculateAngles, nOctaves, nOctaveLayers, firstOctave, angleMode );
-}
-
-void SiftDescriptorExtractor::write (FileStorage &fs) const
-{
-// fs << "algorithm" << getAlgorithmName ();
-
- SIFT::CommonParams commParams = sift.getCommonParams ();
- SIFT::DescriptorParams descriptorParams = sift.getDescriptorParams ();
- fs << "magnification" << descriptorParams.magnification;
- fs << "isNormalize" << descriptorParams.isNormalize;
- fs << "recalculateAngles" << descriptorParams.recalculateAngles;
- fs << "nOctaves" << commParams.nOctaves;
- fs << "nOctaveLayers" << commParams.nOctaveLayers;
- fs << "firstOctave" << commParams.firstOctave;
- fs << "angleMode" << commParams.angleMode;
-}
-
-/****************************************************************************************\
-* SurfDescriptorExtractor *
-\****************************************************************************************/
-SurfDescriptorExtractor::SurfDescriptorExtractor( int nOctaves,
- int nOctaveLayers, bool extended )
- : surf( 0.0, nOctaves, nOctaveLayers, extended )
-{}
-
-void SurfDescriptorExtractor::compute( const Mat& image,
- vector<KeyPoint>& keypoints,
- Mat& descriptors) const
-{
- // Compute descriptors for given keypoints
- vector<float> _descriptors;
- Mat mask;
- bool useProvidedKeypoints = true;
- surf(image, mask, keypoints, _descriptors, useProvidedKeypoints);
-
- descriptors.create((int)keypoints.size(), (int)surf.descriptorSize(), CV_32FC1);
- assert( (int)_descriptors.size() == descriptors.rows * descriptors.cols );
- std::copy(_descriptors.begin(), _descriptors.end(), descriptors.begin<float>());
-}
-
-void SurfDescriptorExtractor::read( const FileNode &fn )
-{
- int nOctaves = fn["nOctaves"];
- int nOctaveLayers = fn["nOctaveLayers"];
- bool extended = (int)fn["extended"] != 0;
-
- surf = SURF( 0.0, nOctaves, nOctaveLayers, extended );
-}
-
-void SurfDescriptorExtractor::write( FileStorage &fs ) const
-{
-// fs << "algorithm" << getAlgorithmName ();
-
- fs << "nOctaves" << surf.nOctaves;
- fs << "nOctaveLayers" << surf.nOctaveLayers;
- fs << "extended" << surf.extended;
-}
-
-/****************************************************************************************\
-* Factory functions for descriptor extractor and matcher creating *
-\****************************************************************************************/
-
-Ptr<DescriptorExtractor> createDescriptorExtractor( const string& descriptorExtractorType )
-{
- DescriptorExtractor* de = 0;
- if( !descriptorExtractorType.compare( "SIFT" ) )
- {
- de = new SiftDescriptorExtractor/*( double magnification=SIFT::DescriptorParams::GET_DEFAULT_MAGNIFICATION(),
- bool isNormalize=true, bool recalculateAngles=true,
- int nOctaves=SIFT::CommonParams::DEFAULT_NOCTAVES,
- int nOctaveLayers=SIFT::CommonParams::DEFAULT_NOCTAVE_LAYERS,
- int firstOctave=SIFT::CommonParams::DEFAULT_FIRST_OCTAVE,
- int angleMode=SIFT::CommonParams::FIRST_ANGLE )*/;
- }
- else if( !descriptorExtractorType.compare( "SURF" ) )
- {
- de = new SurfDescriptorExtractor/*( int nOctaves=4, int nOctaveLayers=2, bool extended=false )*/;
- }
- else
+ if( image.empty() || keypoints.empty() )
{
- //CV_Error( CV_StsBadArg, "unsupported descriptor extractor type");
+ descriptors.release();
+ return;
}
- return de;
-}
-Ptr<DescriptorMatcher> createDescriptorMatcher( const string& descriptorMatcherType )
-{
- DescriptorMatcher* dm = 0;
- if( !descriptorMatcherType.compare( "BruteForce" ) )
- {
- dm = new BruteForceMatcher<L2<float> >();
- }
- else if ( !descriptorMatcherType.compare( "BruteForce-L1" ) )
- {
- dm = new BruteForceMatcher<L1<float> >();
- }
- else
- {
- //CV_Error( CV_StsBadArg, "unsupported descriptor matcher type");
- }
+ KeyPointsFilter::runByImageBorder( keypoints, image.size(), 0 );
+ KeyPointsFilter::runByKeypointSize( keypoints, std::numeric_limits<float>::epsilon() );
- return dm;
+ computeImpl( image, keypoints, descriptors );
}
-/****************************************************************************************\
-* BruteForceMatcher L2 specialization *
-\****************************************************************************************/
-template<>
-void BruteForceMatcher<L2<float> >::matchImpl( const Mat& descriptors_1, const Mat& descriptors_2,
- const Mat& /*mask*/, vector<int>& matches ) const
+void DescriptorExtractor::compute( const vector<Mat>& imageCollection, vector<vector<KeyPoint> >& pointCollection, vector<Mat>& descCollection ) const
{
- matches.clear();
- matches.reserve( descriptors_1.rows );
-//TODO: remove _DEBUG if bag 416 fixed
-#if (defined _DEBUG || !defined HAVE_EIGEN2)
- Mat norms;
- cv::reduce( descriptors_2.mul( descriptors_2 ), norms, 1, 0);
- norms = norms.t();
- Mat desc_2t = descriptors_2.t();
- for( int i=0;i<descriptors_1.rows;i++ )
- {
- Mat distances = (-2)*descriptors_1.row(i)*desc_2t;
- distances += norms;
- Point minLoc;
- minMaxLoc ( distances, 0, 0, &minLoc );
- matches.push_back( minLoc.x );
- }
-
-#else
- Eigen::Matrix<float, Eigen::Dynamic, Eigen::Dynamic> desc1t;
- Eigen::Matrix<float, Eigen::Dynamic, Eigen::Dynamic> desc2;
- cv2eigen( descriptors_1.t(), desc1t);
- cv2eigen( descriptors_2, desc2 );
-
- //Eigen::Matrix<float, Eigen::Dynamic, 1> norms = desc2.rowwise().squaredNorm();
- Eigen::Matrix<float, Eigen::Dynamic, 1> norms = desc2.rowwise().squaredNorm() / 2;
- for( int i=0;i<descriptors_1.rows;i++ )
- {
- //Eigen::Matrix<float, Eigen::Dynamic, 1> distances = (-2) * (desc2*desc1t.col(i));
- Eigen::Matrix<float, Eigen::Dynamic, 1> distances = desc2*desc1t.col(i);
-
- //distances += norms;
- distances -= norms;
-
- int idx;
-
- //distances.minCoeff(&idx);
- distances.maxCoeff(&idx);
- matches.push_back( idx );
- }
-#endif
-}
-
-/****************************************************************************************\
-* GenericDescriptorMatch *
-\****************************************************************************************/
-/*
- * KeyPointCollection
- */
-void KeyPointCollection::add( const Mat& _image, const vector<KeyPoint>& _points )
-{
- // update m_start_indices
- if( startIndices.empty() )
- startIndices.push_back(0);
- else
- startIndices.push_back((int)(*startIndices.rbegin() + points.rbegin()->size()));
-
- // add image and keypoints
- images.push_back(_image);
- points.push_back(_points);
-}
-
-KeyPoint KeyPointCollection::getKeyPoint( int index ) const
-{
- size_t i = 0;
- for(; i < startIndices.size() && startIndices[i] <= index; i++);
- i--;
- assert(i < startIndices.size() && (size_t)index - startIndices[i] < points[i].size());
-
- return points[i][index - startIndices[i]];
-}
-
-size_t KeyPointCollection::calcKeypointCount() const
-{
- if( startIndices.empty() )
- return 0;
- return *startIndices.rbegin() + points.rbegin()->size();
-}
-
-void KeyPointCollection::clear()
-{
- images.clear();
- points.clear();
- startIndices.clear();
-}
-
-/*
- * GenericDescriptorMatch
- */
-
-void GenericDescriptorMatch::match( const Mat&, vector<KeyPoint>&, vector<DMatch>& )
-{
-}
-
-void GenericDescriptorMatch::match( const Mat&, vector<KeyPoint>&, vector<vector<DMatch> >&, float )
-{
-}
-
-void GenericDescriptorMatch::add( KeyPointCollection& collection )
-{
- for( size_t i = 0; i < collection.images.size(); i++ )
- add( collection.images[i], collection.points[i] );
-}
-
-void GenericDescriptorMatch::classify( const Mat& image, vector<cv::KeyPoint>& points )
-{
- vector<int> keypointIndices;
- match( image, points, keypointIndices );
-
- // remap keypoint indices to descriptors
- for( size_t i = 0; i < keypointIndices.size(); i++ )
- points[i].class_id = collection.getKeyPoint(keypointIndices[i]).class_id;
-};
-
-void GenericDescriptorMatch::clear()
-{
- collection.clear();
+ CV_Assert( imageCollection.size() == pointCollection.size() );
+ descCollection.resize( imageCollection.size() );
+ for( size_t i = 0; i < imageCollection.size(); i++ )
+ compute( imageCollection[i], pointCollection[i], descCollection[i] );
}
-/*
- * Factory function for GenericDescriptorMatch creating
- */
-Ptr<GenericDescriptorMatch> createGenericDescriptorMatch( const string& genericDescritptorMatchType, const string ¶msFilename )
-{
- GenericDescriptorMatch *descriptorMatch = 0;
- if( ! genericDescritptorMatchType.compare ("ONEWAY") )
- {
- descriptorMatch = new OneWayDescriptorMatch ();
- }
- else if( ! genericDescritptorMatchType.compare ("FERN") )
- {
- FernDescriptorMatch::Params params;
- params.signatureSize = numeric_limits<int>::max();
- descriptorMatch = new FernDescriptorMatch (params);
- }
- else if( ! genericDescritptorMatchType.compare ("CALONDER") )
- {
- //descriptorMatch = new CalonderDescriptorMatch ();
- }
-
- if( !paramsFilename.empty() && descriptorMatch != 0 )
- {
- FileStorage fs = FileStorage( paramsFilename, FileStorage::READ );
- if( fs.isOpened() )
- {
- descriptorMatch->read( fs.root() );
- fs.release();
- }
- }
-
- return descriptorMatch;
-}
-
-/****************************************************************************************\
-* OneWayDescriptorMatch *
-\****************************************************************************************/
-OneWayDescriptorMatch::OneWayDescriptorMatch()
+/*void DescriptorExtractor::read( const FileNode& )
{}
-OneWayDescriptorMatch::OneWayDescriptorMatch( const Params& _params)
-{
- initialize(_params);
-}
-
-OneWayDescriptorMatch::~OneWayDescriptorMatch()
-{}
-
-void OneWayDescriptorMatch::initialize( const Params& _params, OneWayDescriptorBase *_base)
-{
- base.release();
- if (_base != 0)
- {
- base = _base;
- }
- params = _params;
-}
-
-void OneWayDescriptorMatch::add( const Mat& image, vector<KeyPoint>& keypoints )
-{
- if( base.empty() )
- base = new OneWayDescriptorObject( params.patchSize, params.poseCount, params.pcaFilename,
- params.trainPath, params.trainImagesList, params.minScale, params.maxScale, params.stepScale);
-
- size_t trainFeatureCount = keypoints.size();
-
- base->Allocate( (int)trainFeatureCount );
-
- IplImage _image = image;
- for( size_t i = 0; i < keypoints.size(); i++ )
- base->InitializeDescriptor( (int)i, &_image, keypoints[i], "" );
-
- collection.add( Mat(), keypoints );
-
-#if defined(_KDTREE)
- base->ConvertDescriptorsArrayToTree();
-#endif
-}
-
-void OneWayDescriptorMatch::add( KeyPointCollection& keypoints )
-{
- if( base.empty() )
- base = new OneWayDescriptorObject( params.patchSize, params.poseCount, params.pcaFilename,
- params.trainPath, params.trainImagesList, params.minScale, params.maxScale, params.stepScale);
-
- size_t trainFeatureCount = keypoints.calcKeypointCount();
-
- base->Allocate( (int)trainFeatureCount );
-
- int count = 0;
- for( size_t i = 0; i < keypoints.points.size(); i++ )
- {
- for( size_t j = 0; j < keypoints.points[i].size(); j++ )
- {
- IplImage img = keypoints.images[i];
- base->InitializeDescriptor( count++, &img, keypoints.points[i][j], "" );
- }
+void DescriptorExtractor::write( FileStorage& ) const
+{}*/
- collection.add( Mat(), keypoints.points[i] );
- }
-
-#if defined(_KDTREE)
- base->ConvertDescriptorsArrayToTree();
-#endif
-}
-
-void OneWayDescriptorMatch::match( const Mat& image, vector<KeyPoint>& points, vector<int>& indices)
+bool DescriptorExtractor::empty() const
{
- vector<DMatch> matchings( points.size() );
- indices.resize(points.size());
-
- match( image, points, matchings );
-
- for( size_t i = 0; i < points.size(); i++ )
- indices[i] = matchings[i].indexTrain;
+ return false;
}
-void OneWayDescriptorMatch::match( const Mat& image, vector<KeyPoint>& points, vector<DMatch>& matches )
+void DescriptorExtractor::removeBorderKeypoints( vector<KeyPoint>& keypoints,
+ Size imageSize, int borderSize )
{
- matches.resize( points.size() );
- IplImage _image = image;
- for( size_t i = 0; i < points.size(); i++ )
- {
- int poseIdx = -1;
-
- DMatch match;
- match.indexQuery = (int)i;
- match.indexTrain = -1;
- base->FindDescriptor( &_image, points[i].pt, match.indexTrain, poseIdx, match.distance );
- matches[i] = match;
- }
+ KeyPointsFilter::runByImageBorder( keypoints, imageSize, borderSize );
}
-void OneWayDescriptorMatch::match( const Mat& image, vector<KeyPoint>& points, vector<vector<DMatch> >& matches, float /*threshold*/ )
+Ptr<DescriptorExtractor> DescriptorExtractor::create(const string& descriptorExtractorType)
{
- matches.clear();
- matches.resize( points.size() );
-
- vector<DMatch> dmatches;
- match( image, points, dmatches );
- for( size_t i=0;i<matches.size();i++ )
- {
- matches[i].push_back( dmatches[i] );
- }
-
- /*
- printf("Start matching %d points\n", points.size());
- //std::cout << "Start matching " << points.size() << "points\n";
- assert(collection.images.size() == 1);
- int n = collection.points[0].size();
-
- printf("n = %d\n", n);
- for( size_t i = 0; i < points.size(); i++ )
+ if( descriptorExtractorType.find("Opponent") == 0 )
{
- //printf("Matching %d\n", i);
- //int poseIdx = -1;
-
- DMatch match;
- match.indexQuery = i;
- match.indexTrain = -1;
-
-
- CvPoint pt = points[i].pt;
- CvRect roi = cvRect(cvRound(pt.x - 24/4),
- cvRound(pt.y - 24/4),
- 24/2, 24/2);
- cvSetImageROI(&_image, roi);
-
- std::vector<int> desc_idxs;
- std::vector<int> pose_idxs;
- std::vector<float> distances;
- std::vector<float> _scales;
-
-
- base->FindDescriptor(&_image, n, desc_idxs, pose_idxs, distances, _scales);
- cvResetImageROI(&_image);
-
- for( int j=0;j<n;j++ )
- {
- match.indexTrain = desc_idxs[j];
- match.distance = distances[j];
- matches[i].push_back( match );
- }
-
- //sort( matches[i].begin(), matches[i].end(), compareIndexTrain );
- //for( int j=0;j<n;j++ )
- //{
- //printf( "%d %f; ",matches[i][j].indexTrain, matches[i][j].distance);
- //}
- //printf("\n\n\n");
-
-
-
- //base->FindDescriptor( &_image, 100, points[i].pt, match.indexTrain, poseIdx, match.distance );
- //matches[i].push_back( match );
+ size_t pos = string("Opponent").size();
+ string type = descriptorExtractorType.substr(pos);
+ return new OpponentColorDescriptorExtractor(DescriptorExtractor::create(type));
}
- */
-}
-
-void OneWayDescriptorMatch::read( const FileNode &fn )
-{
- base = new OneWayDescriptorObject( params.patchSize, params.poseCount, string (), string (), string (),
- params.minScale, params.maxScale, params.stepScale );
- base->Read (fn);
-}
-
-
-void OneWayDescriptorMatch::write( FileStorage& fs ) const
-{
- base->Write (fs);
+ return Algorithm::create<DescriptorExtractor>("Feature2D." + descriptorExtractorType);
}
-void OneWayDescriptorMatch::classify( const Mat& image, vector<KeyPoint>& points )
-{
- IplImage _image = image;
- for( size_t i = 0; i < points.size(); i++ )
- {
- int descIdx = -1;
- int poseIdx = -1;
- float distance;
- base->FindDescriptor(&_image, points[i].pt, descIdx, poseIdx, distance);
- points[i].class_id = collection.getKeyPoint(descIdx).class_id;
- }
-}
-
-void OneWayDescriptorMatch::clear ()
-{
- GenericDescriptorMatch::clear();
- base->clear ();
-}
+/////////////////////////////////////////////////////////////////////////////////////////////////////////////////
/****************************************************************************************\
-* CalonderDescriptorMatch *
+* OpponentColorDescriptorExtractor *
\****************************************************************************************/
-#if 0
-CalonderDescriptorMatch::Params::Params( const RNG& _rng, const PatchGenerator& _patchGen,
- int _numTrees, int _depth, int _views,
- size_t _reducedNumDim,
- int _numQuantBits,
- bool _printStatus,
- int _patchSize ) :
- rng(_rng), patchGen(_patchGen), numTrees(_numTrees), depth(_depth), views(_views),
- patchSize(_patchSize), reducedNumDim(_reducedNumDim), numQuantBits(_numQuantBits), printStatus(_printStatus)
-{}
-
-CalonderDescriptorMatch::Params::Params( const string& _filename )
+OpponentColorDescriptorExtractor::OpponentColorDescriptorExtractor( const Ptr<DescriptorExtractor>& _descriptorExtractor ) :
+ descriptorExtractor(_descriptorExtractor)
{
- filename = _filename;
+ CV_Assert( !descriptorExtractor.empty() );
}
-CalonderDescriptorMatch::CalonderDescriptorMatch()
-{}
-
-CalonderDescriptorMatch::CalonderDescriptorMatch( const Params& _params )
+static void convertBGRImageToOpponentColorSpace( const Mat& bgrImage, vector<Mat>& opponentChannels )
{
- initialize(_params);
-}
+ if( bgrImage.type() != CV_8UC3 )
+ CV_Error( CV_StsBadArg, "input image must be an BGR image of type CV_8UC3" );
-CalonderDescriptorMatch::~CalonderDescriptorMatch()
-{}
+ // Split image into RGB to allow conversion to Opponent Color Space.
+ vector<Mat> bgrChannels(3);
+ split( bgrImage, bgrChannels );
-void CalonderDescriptorMatch::initialize( const Params& _params )
-{
- classifier.release();
- params = _params;
- if( !params.filename.empty() )
- {
- classifier = new RTreeClassifier;
- classifier->read( params.filename.c_str() );
- }
-}
-
-void CalonderDescriptorMatch::add( const Mat& image, vector<KeyPoint>& keypoints )
-{
- if( params.filename.empty() )
- collection.add( image, keypoints );
-}
-
-Mat CalonderDescriptorMatch::extractPatch( const Mat& image, const Point& pt, int patchSize ) const
-{
- const int offset = patchSize / 2;
- return image( Rect(pt.x - offset, pt.y - offset, patchSize, patchSize) );
-}
-
-void CalonderDescriptorMatch::calcBestProbAndMatchIdx( const Mat& image, const Point& pt,
- float& bestProb, int& bestMatchIdx, float* signature )
-{
- IplImage roi = extractPatch( image, pt, params.patchSize );
- classifier->getSignature( &roi, signature );
+ // Prepare opponent color space storage matrices.
+ opponentChannels.resize( 3 );
+ opponentChannels[0] = cv::Mat(bgrImage.size(), CV_8UC1); // R-G RED-GREEN
+ opponentChannels[1] = cv::Mat(bgrImage.size(), CV_8UC1); // R+G-2B YELLOW-BLUE
+ opponentChannels[2] = cv::Mat(bgrImage.size(), CV_8UC1); // R+G+B
- bestProb = 0;
- bestMatchIdx = -1;
- for( int ci = 0; ci < classifier->classes(); ci++ )
+ // Calculate the channels of the opponent color space
{
- if( signature[ci] > bestProb )
+ // (R - G) / sqrt(2)
+ MatConstIterator_<signed char> rIt = bgrChannels[2].begin<signed char>();
+ MatConstIterator_<signed char> gIt = bgrChannels[1].begin<signed char>();
+ MatIterator_<unsigned char> dstIt = opponentChannels[0].begin<unsigned char>();
+ float factor = 1.f / sqrt(2.f);
+ for( ; dstIt != opponentChannels[0].end<unsigned char>(); ++rIt, ++gIt, ++dstIt )
{
- bestProb = signature[ci];
- bestMatchIdx = ci;
+ int value = static_cast<int>( static_cast<float>(static_cast<int>(*gIt)-static_cast<int>(*rIt)) * factor );
+ if( value < 0 ) value = 0;
+ if( value > 255 ) value = 255;
+ (*dstIt) = static_cast<unsigned char>(value);
}
}
-}
-
-void CalonderDescriptorMatch::trainRTreeClassifier()
-{
- if( classifier.empty() )
{
- assert( params.filename.empty() );
- classifier = new RTreeClassifier;
-
- vector<BaseKeypoint> baseKeyPoints;
- vector<IplImage> iplImages( collection.images.size() );
- for( size_t imageIdx = 0; imageIdx < collection.images.size(); imageIdx++ )
+ // (R + G - 2B)/sqrt(6)
+ MatConstIterator_<signed char> rIt = bgrChannels[2].begin<signed char>();
+ MatConstIterator_<signed char> gIt = bgrChannels[1].begin<signed char>();
+ MatConstIterator_<signed char> bIt = bgrChannels[0].begin<signed char>();
+ MatIterator_<unsigned char> dstIt = opponentChannels[1].begin<unsigned char>();
+ float factor = 1.f / sqrt(6.f);
+ for( ; dstIt != opponentChannels[1].end<unsigned char>(); ++rIt, ++gIt, ++bIt, ++dstIt )
{
- iplImages[imageIdx] = collection.images[imageIdx];
- for( size_t pointIdx = 0; pointIdx < collection.points[imageIdx].size(); pointIdx++ )
- {
- BaseKeypoint bkp;
- KeyPoint kp = collection.points[imageIdx][pointIdx];
- bkp.x = cvRound(kp.pt.x);
- bkp.y = cvRound(kp.pt.y);
- bkp.image = &iplImages[imageIdx];
- baseKeyPoints.push_back(bkp);
- }
+ int value = static_cast<int>( static_cast<float>(static_cast<int>(*rIt) + static_cast<int>(*gIt) - 2*static_cast<int>(*bIt)) *
+ factor );
+ if( value < 0 ) value = 0;
+ if( value > 255 ) value = 255;
+ (*dstIt) = static_cast<unsigned char>(value);
}
- classifier->train( baseKeyPoints, params.rng, params.patchGen, params.numTrees,
- params.depth, params.views, params.reducedNumDim, params.numQuantBits,
- params.printStatus );
}
-}
-
-void CalonderDescriptorMatch::match( const Mat& image, vector<KeyPoint>& keypoints, vector<int>& indices )
-{
- trainRTreeClassifier();
-
- float bestProb = 0;
- AutoBuffer<float> signature( classifier->classes() );
- indices.resize( keypoints.size() );
-
- for( size_t pi = 0; pi < keypoints.size(); pi++ )
- calcBestProbAndMatchIdx( image, keypoints[pi].pt, bestProb, indices[pi], signature );
-}
-
-void CalonderDescriptorMatch::classify( const Mat& image, vector<KeyPoint>& keypoints )
-{
- trainRTreeClassifier();
-
- AutoBuffer<float> signature( classifier->classes() );
- for( size_t pi = 0; pi < keypoints.size(); pi++ )
{
- float bestProb = 0;
- int bestMatchIdx = -1;
- calcBestProbAndMatchIdx( image, keypoints[pi].pt, bestProb, bestMatchIdx, signature );
- keypoints[pi].class_id = collection.getKeyPoint(bestMatchIdx).class_id;
- }
-}
-
-void CalonderDescriptorMatch::clear ()
-{
- GenericDescriptorMatch::clear();
- classifier.release();
-}
-
-void CalonderDescriptorMatch::read( const FileNode &fn )
-{
- params.numTrees = fn["numTrees"];
- params.depth = fn["depth"];
- params.views = fn["views"];
- params.patchSize = fn["patchSize"];
- params.reducedNumDim = (int) fn["reducedNumDim"];
- params.numQuantBits = fn["numQuantBits"];
- params.printStatus = (int) fn["printStatus"] != 0;
-}
-
-void CalonderDescriptorMatch::write( FileStorage& fs ) const
-{
- fs << "numTrees" << params.numTrees;
- fs << "depth" << params.depth;
- fs << "views" << params.views;
- fs << "patchSize" << params.patchSize;
- fs << "reducedNumDim" << (int) params.reducedNumDim;
- fs << "numQuantBits" << params.numQuantBits;
- fs << "printStatus" << params.printStatus;
-}
-#endif
-
-/****************************************************************************************\
-* FernDescriptorMatch *
-\****************************************************************************************/
-FernDescriptorMatch::Params::Params( int _nclasses, int _patchSize, int _signatureSize,
- int _nstructs, int _structSize, int _nviews, int _compressionMethod,
- const PatchGenerator& _patchGenerator ) :
- nclasses(_nclasses), patchSize(_patchSize), signatureSize(_signatureSize),
- nstructs(_nstructs), structSize(_structSize), nviews(_nviews),
- compressionMethod(_compressionMethod), patchGenerator(_patchGenerator)
-{}
-
-FernDescriptorMatch::Params::Params( const string& _filename )
-{
- filename = _filename;
-}
-
-FernDescriptorMatch::FernDescriptorMatch()
-{}
-
-FernDescriptorMatch::FernDescriptorMatch( const Params& _params )
-{
- params = _params;
-}
-
-FernDescriptorMatch::~FernDescriptorMatch()
-{}
-
-void FernDescriptorMatch::initialize( const Params& _params )
-{
- classifier.release();
- params = _params;
- if( !params.filename.empty() )
- {
- classifier = new FernClassifier;
- FileStorage fs(params.filename, FileStorage::READ);
- if( fs.isOpened() )
- classifier->read( fs.getFirstTopLevelNode() );
- }
-}
-
-void FernDescriptorMatch::add( const Mat& image, vector<KeyPoint>& keypoints )
-{
- if( params.filename.empty() )
- collection.add( image, keypoints );
-}
-
-void FernDescriptorMatch::trainFernClassifier()
-{
- if( classifier.empty() )
- {
- assert( params.filename.empty() );
-
- vector<vector<Point2f> > points;
- for( size_t imgIdx = 0; imgIdx < collection.images.size(); imgIdx++ )
- KeyPoint::convert( collection.points[imgIdx], points[imgIdx] );
-
- classifier = new FernClassifier( points, collection.images, vector<vector<int> >(), 0, // each points is a class
- params.patchSize, params.signatureSize, params.nstructs, params.structSize,
- params.nviews, params.compressionMethod, params.patchGenerator );
- }
-}
-
-void FernDescriptorMatch::calcBestProbAndMatchIdx( const Mat& image, const Point2f& pt,
- float& bestProb, int& bestMatchIdx, vector<float>& signature )
-{
- (*classifier)( image, pt, signature);
-
- bestProb = -FLT_MAX;
- bestMatchIdx = -1;
- for( int ci = 0; ci < classifier->getClassCount(); ci++ )
- {
- if( signature[ci] > bestProb )
+ // (R + G + B)/sqrt(3)
+ MatConstIterator_<signed char> rIt = bgrChannels[2].begin<signed char>();
+ MatConstIterator_<signed char> gIt = bgrChannels[1].begin<signed char>();
+ MatConstIterator_<signed char> bIt = bgrChannels[0].begin<signed char>();
+ MatIterator_<unsigned char> dstIt = opponentChannels[2].begin<unsigned char>();
+ float factor = 1.f / sqrt(3.f);
+ for( ; dstIt != opponentChannels[2].end<unsigned char>(); ++rIt, ++gIt, ++bIt, ++dstIt )
{
- bestProb = signature[ci];
- bestMatchIdx = ci;
+ int value = static_cast<int>( static_cast<float>(static_cast<int>(*rIt) + static_cast<int>(*gIt) + static_cast<int>(*bIt)) *
+ factor );
+ if( value < 0 ) value = 0;
+ if( value > 255 ) value = 255;
+ (*dstIt) = static_cast<unsigned char>(value);
}
}
}
-void FernDescriptorMatch::match( const Mat& image, vector<KeyPoint>& keypoints, vector<int>& indices )
-{
- trainFernClassifier();
-
- indices.resize( keypoints.size() );
- vector<float> signature( (size_t)classifier->getClassCount() );
-
- for( size_t pi = 0; pi < keypoints.size(); pi++ )
- {
- //calcBestProbAndMatchIdx( image, keypoints[pi].pt, bestProb, indices[pi], signature );
- //TODO: use octave and image pyramid
- indices[pi] = (*classifier)(image, keypoints[pi].pt, signature);
- }
-}
-
-void FernDescriptorMatch::match( const Mat& image, vector<KeyPoint>& keypoints, vector<DMatch>& matches )
+struct KP_LessThan
{
- trainFernClassifier();
-
- matches.resize( keypoints.size() );
- vector<float> signature( (size_t)classifier->getClassCount() );
-
- for( int pi = 0; pi < (int)keypoints.size(); pi++ )
+ KP_LessThan(const vector<KeyPoint>& _kp) : kp(&_kp) {}
+ bool operator()(int i, int j) const
{
- matches[pi].indexQuery = pi;
- calcBestProbAndMatchIdx( image, keypoints[pi].pt, matches[pi].distance, matches[pi].indexTrain, signature );
- //matching[pi].distance is log of probability so we need to transform it
- matches[pi].distance = -matches[pi].distance;
+ return (*kp)[i].class_id < (*kp)[j].class_id;
}
-}
+ const vector<KeyPoint>* kp;
+};
-void FernDescriptorMatch::match( const Mat& image, vector<KeyPoint>& keypoints, vector<vector<DMatch> >& matches, float threshold )
+void OpponentColorDescriptorExtractor::computeImpl( const Mat& bgrImage, vector<KeyPoint>& keypoints, Mat& descriptors ) const
{
- trainFernClassifier();
+ vector<Mat> opponentChannels;
+ convertBGRImageToOpponentColorSpace( bgrImage, opponentChannels );
- matches.resize( keypoints.size() );
- vector<float> signature( (size_t)classifier->getClassCount() );
+ const int N = 3; // channels count
+ vector<KeyPoint> channelKeypoints[N];
+ Mat channelDescriptors[N];
+ vector<int> idxs[N];
- for( int pi = 0; pi < (int)keypoints.size(); pi++ )
+ // Compute descriptors three times, once for each Opponent channel to concatenate into a single color descriptor
+ int maxKeypointsCount = 0;
+ for( int ci = 0; ci < N; ci++ )
{
- (*classifier)( image, keypoints[pi].pt, signature);
+ channelKeypoints[ci].insert( channelKeypoints[ci].begin(), keypoints.begin(), keypoints.end() );
+ // Use class_id member to get indices into initial keypoints vector
+ for( size_t ki = 0; ki < channelKeypoints[ci].size(); ki++ )
+ channelKeypoints[ci][ki].class_id = (int)ki;
- DMatch match;
- match.indexQuery = pi;
-
- for( int ci = 0; ci < classifier->getClassCount(); ci++ )
+ descriptorExtractor->compute( opponentChannels[ci], channelKeypoints[ci], channelDescriptors[ci] );
+ idxs[ci].resize( channelKeypoints[ci].size() );
+ for( size_t ki = 0; ki < channelKeypoints[ci].size(); ki++ )
{
- if( -signature[ci] < threshold )
+ idxs[ci][ki] = (int)ki;
+ }
+ std::sort( idxs[ci].begin(), idxs[ci].end(), KP_LessThan(channelKeypoints[ci]) );
+ maxKeypointsCount = std::max( maxKeypointsCount, (int)channelKeypoints[ci].size());
+ }
+
+ vector<KeyPoint> outKeypoints;
+ outKeypoints.reserve( keypoints.size() );
+
+ int dSize = descriptorExtractor->descriptorSize();
+ Mat mergedDescriptors( maxKeypointsCount, 3*dSize, descriptorExtractor->descriptorType() );
+ int mergedCount = 0;
+ // cp - current channel position
+ size_t cp[] = {0, 0, 0};
+ while( cp[0] < channelKeypoints[0].size() &&
+ cp[1] < channelKeypoints[1].size() &&
+ cp[2] < channelKeypoints[2].size() )
+ {
+ const int maxInitIdx = std::max( 0, std::max( channelKeypoints[0][idxs[0][cp[0]]].class_id,
+ std::max( channelKeypoints[1][idxs[1][cp[1]]].class_id,
+ channelKeypoints[2][idxs[2][cp[2]]].class_id ) ) );
+
+ while( channelKeypoints[0][idxs[0][cp[0]]].class_id < maxInitIdx && cp[0] < channelKeypoints[0].size() ) { cp[0]++; }
+ while( channelKeypoints[1][idxs[1][cp[1]]].class_id < maxInitIdx && cp[1] < channelKeypoints[1].size() ) { cp[1]++; }
+ while( channelKeypoints[2][idxs[2][cp[2]]].class_id < maxInitIdx && cp[2] < channelKeypoints[2].size() ) { cp[2]++; }
+ if( cp[0] >= channelKeypoints[0].size() || cp[1] >= channelKeypoints[1].size() || cp[2] >= channelKeypoints[2].size() )
+ break;
+
+ if( channelKeypoints[0][idxs[0][cp[0]]].class_id == maxInitIdx &&
+ channelKeypoints[1][idxs[1][cp[1]]].class_id == maxInitIdx &&
+ channelKeypoints[2][idxs[2][cp[2]]].class_id == maxInitIdx )
+ {
+ outKeypoints.push_back( keypoints[maxInitIdx] );
+ // merge descriptors
+ for( int ci = 0; ci < N; ci++ )
{
- match.distance = -signature[ci];
- match.indexTrain = ci;
- matches[pi].push_back( match );
+ Mat dst = mergedDescriptors(Range(mergedCount, mergedCount+1), Range(ci*dSize, (ci+1)*dSize));
+ channelDescriptors[ci].row( idxs[ci][cp[ci]] ).copyTo( dst );
+ cp[ci]++;
}
+ mergedCount++;
}
}
+ mergedDescriptors.rowRange(0, mergedCount).copyTo( descriptors );
+ std::swap( outKeypoints, keypoints );
}
-void FernDescriptorMatch::classify( const Mat& image, vector<KeyPoint>& keypoints )
-{
- trainFernClassifier();
-
- vector<float> signature( (size_t)classifier->getClassCount() );
- for( size_t pi = 0; pi < keypoints.size(); pi++ )
- {
- float bestProb = 0;
- int bestMatchIdx = -1;
- calcBestProbAndMatchIdx( image, keypoints[pi].pt, bestProb, bestMatchIdx, signature );
- keypoints[pi].class_id = collection.getKeyPoint(bestMatchIdx).class_id;
- }
-}
-
-void FernDescriptorMatch::read( const FileNode &fn )
+void OpponentColorDescriptorExtractor::read( const FileNode& fn )
{
- params.nclasses = fn["nclasses"];
- params.patchSize = fn["patchSize"];
- params.signatureSize = fn["signatureSize"];
- params.nstructs = fn["nstructs"];
- params.structSize = fn["structSize"];
- params.nviews = fn["nviews"];
- params.compressionMethod = fn["compressionMethod"];
-
- //classifier->read(fn);
+ descriptorExtractor->read(fn);
}
-void FernDescriptorMatch::write( FileStorage& fs ) const
+void OpponentColorDescriptorExtractor::write( FileStorage& fs ) const
{
- fs << "nclasses" << params.nclasses;
- fs << "patchSize" << params.patchSize;
- fs << "signatureSize" << params.signatureSize;
- fs << "nstructs" << params.nstructs;
- fs << "structSize" << params.structSize;
- fs << "nviews" << params.nviews;
- fs << "compressionMethod" << params.compressionMethod;
-
-// classifier->write(fs);
-}
-
-void FernDescriptorMatch::clear ()
-{
- GenericDescriptorMatch::clear();
- classifier.release();
-}
-
-/****************************************************************************************\
-* VectorDescriptorMatch *
-\****************************************************************************************/
-void VectorDescriptorMatch::add( const Mat& image, vector<KeyPoint>& keypoints )
-{
- Mat descriptors;
- extractor->compute( image, keypoints, descriptors );
- matcher->add( descriptors );
-
- collection.add( Mat(), keypoints );
-};
-
-void VectorDescriptorMatch::match( const Mat& image, vector<KeyPoint>& points, vector<int>& keypointIndices )
-{
- Mat descriptors;
- extractor->compute( image, points, descriptors );
-
- matcher->match( descriptors, keypointIndices );
-};
-
-void VectorDescriptorMatch::match( const Mat& image, vector<KeyPoint>& points, vector<DMatch>& matches )
-{
- Mat descriptors;
- extractor->compute( image, points, descriptors );
-
- matcher->match( descriptors, matches );
-}
-
-void VectorDescriptorMatch::match( const Mat& image, vector<KeyPoint>& points,
- vector<vector<DMatch> >& matches, float threshold )
-{
- Mat descriptors;
- extractor->compute( image, points, descriptors );
-
- matcher->match( descriptors, matches, threshold );
+ descriptorExtractor->write(fs);
}
-void VectorDescriptorMatch::clear()
+int OpponentColorDescriptorExtractor::descriptorSize() const
{
- GenericDescriptorMatch::clear();
- matcher->clear();
+ return 3*descriptorExtractor->descriptorSize();
}
-void VectorDescriptorMatch::read( const FileNode& fn )
+int OpponentColorDescriptorExtractor::descriptorType() const
{
- GenericDescriptorMatch::read(fn);
- extractor->read (fn);
+ return descriptorExtractor->descriptorType();
}
-void VectorDescriptorMatch::write (FileStorage& fs) const
+bool OpponentColorDescriptorExtractor::empty() const
{
- GenericDescriptorMatch::write(fs);
- extractor->write (fs);
+ return descriptorExtractor.empty() || (DescriptorExtractor*)(descriptorExtractor)->empty();
}
}