1 \section{Feature Detection}
6 Implements the Canny algorithm for edge detection.
9 void cvCanny(\par const CvArr* image,
11 \par double threshold1,
12 \par double threshold2,
13 \par int aperture\_size=3 );
14 }\cvdefPy{Canny(image,edges,threshold1,threshold2,aperture\_size=3)-> None}
16 \cvarg{image}{Single-channel input image}
17 \cvarg{edges}{Single-channel image to store the edges found by the function}
18 \cvarg{threshold1}{The first threshold}
19 \cvarg{threshold2}{The second threshold}
20 \cvarg{aperture\_size}{Aperture parameter for the Sobel operator (see \cvCPyCross{Sobel})}
23 The function finds the edges on the input image \texttt{image} and marks them in the output image \texttt{edges} using the Canny algorithm. The smallest value between \texttt{threshold1} and \texttt{threshold2} is used for edge linking, the largest value is used to find the initial segments of strong edges.
25 \cvCPyFunc{CornerEigenValsAndVecs}
26 Calculates eigenvalues and eigenvectors of image blocks for corner detection.
29 void cvCornerEigenValsAndVecs( \par const CvArr* image,\par CvArr* eigenvv,\par int blockSize,\par int aperture\_size=3 );
31 }\cvdefPy{CornerEigenValsAndVecs(image,eigenvv,blockSize,aperture\_size=3)-> None}
34 \cvarg{image}{Input image}
35 \cvarg{eigenvv}{Image to store the results. It must be 6 times wider than the input image}
36 \cvarg{blockSize}{Neighborhood size (see discussion)}
37 \cvarg{aperture\_size}{Aperture parameter for the Sobel operator (see \cvCPyCross{Sobel})}
40 For every pixel, the function \texttt{cvCornerEigenValsAndVecs} considers a $\texttt{blockSize} \times \texttt{blockSize}$ neigborhood S(p). It calcualtes the covariation matrix of derivatives over the neigborhood as:
44 \sum_{S(p)}(dI/dx)^2 & \sum_{S(p)}(dI/dx \cdot dI/dy)^2 \\
45 \sum_{S(p)}(dI/dx \cdot dI/dy)^2 & \sum_{S(p)}(dI/dy)^2
49 After that it finds eigenvectors and eigenvalues of the matrix and stores them into destination image in form
50 $(\lambda_1, \lambda_2, x_1, y_1, x_2, y_2)$ where
52 \item[$\lambda_1, \lambda_2$]are the eigenvalues of $M$; not sorted
53 \item[$x_1, y_1$]are the eigenvectors corresponding to $\lambda_1$
54 \item[$x_2, y_2$]are the eigenvectors corresponding to $\lambda_2$
57 \cvCPyFunc{CornerHarris}
62 \par const CvArr* image,
63 \par CvArr* harris\_dst,
65 \par int aperture\_size=3,
68 \cvdefPy{CornerHarris(image,harris\_dst,blockSize,aperture\_size=3,k=0.04)-> None}
71 \cvarg{image}{Input image}
72 \cvarg{harris\_dst}{Image to store the Harris detector responses. Should have the same size as \texttt{image}}
73 \cvarg{blockSize}{Neighborhood size (see the discussion of \cvCPyCross{CornerEigenValsAndVecs})}
74 \cvarg{aperture\_size}{Aperture parameter for the Sobel operator (see \cvCPyCross{Sobel}).}
75 % format. In the case of floating-point input format this parameter is the number of the fixed float filter used for differencing
76 \cvarg{k}{Harris detector free parameter. See the formula below}
79 The function runs the Harris edge detector on the image. Similarly to \cvCPyCross{CornerMinEigenVal} and \cvCPyCross{CornerEigenValsAndVecs}, for each pixel it calculates a $2\times2$ gradient covariation matrix $M$ over a $\texttt{blockSize} \times \texttt{blockSize}$ neighborhood. Then, it stores
82 det(M) - k \, trace(M)^2
85 to the destination image. Corners in the image can be found as the local maxima of the destination image.
87 \cvCPyFunc{CornerMinEigenVal}
88 Calculates the minimal eigenvalue of gradient matrices for corner detection.
91 void cvCornerMinEigenVal(
92 \par const CvArr* image,
95 \par int aperture\_size=3 );
96 }\cvdefPy{CornerMinEigenVal(image,eigenval,blockSize,aperture\_size=3)-> None}
98 \cvarg{image}{Input image}
99 \cvarg{eigenval}{Image to store the minimal eigenvalues. Should have the same size as \texttt{image}}
100 \cvarg{blockSize}{Neighborhood size (see the discussion of \cvCPyCross{CornerEigenValsAndVecs})}
101 \cvarg{aperture\_size}{Aperture parameter for the Sobel operator (see \cvCPyCross{Sobel}).}
102 % format. In the case of floating-point input format this parameter is the number of the fixed float filter used for differencing
105 The function is similar to \cvCPyCross{CornerEigenValsAndVecs} but it calculates and stores only the minimal eigen value of derivative covariation matrix for every pixel, i.e. $min(\lambda_1, \lambda_2)$ in terms of the previous function.
107 \cvCPyFunc{ExtractSURF}
108 Extracts Speeded Up Robust Features from an image.
111 void cvExtractSURF( \par const CvArr* image,\par const CvArr* mask,\par CvSeq** keypoints,\par CvSeq** descriptors,\par CvMemStorage* storage,\par CvSURFParams params );
113 \cvdefPy{ExtractSURF(image,mask,storage,params)-> (keypoints,descriptors)}
116 \cvarg{image}{The input 8-bit grayscale image}
117 \cvarg{mask}{The optional input 8-bit mask. The features are only found in the areas that contain more than 50\% of non-zero mask pixels}
118 \cvarg{keypoints}{The output parameter; double pointer to the sequence of keypoints. The sequence of CvSURFPoint structures is as follows:}
120 typedef struct CvSURFPoint
122 CvPoint2D32f pt; // position of the feature within the image
123 int laplacian; // -1, 0 or +1. sign of the laplacian at the point.
124 // can be used to speedup feature comparison
125 // (normally features with laplacians of different
126 // signs can not match)
127 int size; // size of the feature
128 float dir; // orientation of the feature: 0..360 degrees
129 float hessian; // value of the hessian (can be used to
130 // approximately estimate the feature strengths;
131 // see also params.hessianThreshold)
135 \cvarg{descriptors}{The optional output parameter; double pointer to the sequence of descriptors. Depending on the params.extended value, each element of the sequence will be either a 64-element or a 128-element floating-point (\texttt{CV\_32F}) vector. If the parameter is NULL, the descriptors are not computed}
136 \cvarg{storage}{Memory storage where keypoints and descriptors will be stored}
137 \cvarg{params}{Various algorithm parameters put to the structure CvSURFParams:}
139 typedef struct CvSURFParams
141 int extended; // 0 means basic descriptors (64 elements each),
142 // 1 means extended descriptors (128 elements each)
143 double hessianThreshold; // only features with keypoint.hessian
144 // larger than that are extracted.
145 // good default value is ~300-500 (can depend on the
146 // average local contrast and sharpness of the image).
147 // user can further filter out some features based on
148 // their hessian values and other characteristics.
149 int nOctaves; // the number of octaves to be used for extraction.
150 // With each next octave the feature size is doubled
152 int nOctaveLayers; // The number of layers within each octave
157 CvSURFParams cvSURFParams(double hessianThreshold, int extended=0);
158 // returns default parameters
162 The function cvExtractSURF finds robust features in the image, as
165 . For each feature it returns its location, size,
166 orientation and optionally the descriptor, basic or extended. The function
167 can be used for object tracking and localization, image stitching etc. See the
168 \texttt{find\_obj.cpp} demo in OpenCV samples directory.
170 \cvCPyFunc{FindCornerSubPix}
171 Refines the corner locations.
174 void cvFindCornerSubPix(
175 \par const CvArr* image,
176 \par CvPoint2D32f* corners,
179 \par CvSize zero\_zone,
180 \par CvTermCriteria criteria );
181 }\cvdefPy{FindCornerSubPix(image,corners,win,zero\_zone,criteria)-> corners}
184 \cvarg{image}{Input image}
186 \cvarg{corners}{Initial coordinates of the input corners; refined coordinates on output}
187 \cvarg{count}{Number of corners}
190 \cvarg{corners}{Initial coordinates of the input corners as a list of (x, y) pairs}
192 \cvarg{win}{Half of the side length of the search window. For example, if \texttt{win}=(5,5), then a $5*2+1 \times 5*2+1 = 11 \times 11$ search window would be used}
193 \cvarg{zero\_zone}{Half of the size of the dead region in the middle of the search zone over which the summation in the formula below is not done. It is used sometimes to avoid possible singularities of the autocorrelation matrix. The value of (-1,-1) indicates that there is no such size}
194 \cvarg{criteria}{Criteria for termination of the iterative process of corner refinement. That is, the process of corner position refinement stops either after a certain number of iterations or when a required accuracy is achieved. The \texttt{criteria} may specify either of or both the maximum number of iteration and the required accuracy}
197 The function iterates to find the sub-pixel accurate location of corners, or radial saddle points, as shown in on the picture below.
199 It returns the refined coordinates as a list of (x, y) pairs.
202 \includegraphics[width=1.0\textwidth]{pics/cornersubpix.png}
204 Sub-pixel accurate corner locator is based on the observation that every vector from the center $q$ to a point $p$ located within a neighborhood of $q$ is orthogonal to the image gradient at $p$ subject to image and measurement noise. Consider the expression:
207 \epsilon_i = {DI_{p_i}}^T \cdot (q - p_i)
210 where ${DI_{p_i}}$ is the image gradient at the one of the points $p_i$ in a neighborhood of $q$. The value of $q$ is to be found such that $\epsilon_i$ is minimized. A system of equations may be set up with $\epsilon_i$ set to zero:
213 \sum_i(DI_{p_i} \cdot {DI_{p_i}}^T) q = \sum_i(DI_{p_i} \cdot {DI_{p_i}}^T \cdot p_i)
216 where the gradients are summed within a neighborhood ("search window") of $q$. Calling the first gradient term $G$ and the second gradient term $b$ gives:
222 The algorithm sets the center of the neighborhood window at this new center $q$ and then iterates until the center keeps within a set threshold.
224 \cvCPyFunc{GetStarKeypoints}
225 Retrieves keypoints using the StarDetector algorithm.
228 CvSeq* cvGetStarKeypoints( \par const CvArr* image,\par CvMemStorage* storage,\par CvStarDetectorParams params=cvStarDetectorParams() );
229 }\cvdefPy{GetStarKeypoints(image,storage,params)-> keypoints}
232 \cvarg{image}{The input 8-bit grayscale image}
233 \cvarg{storage}{Memory storage where the keypoints will be stored}
234 \cvarg{params}{Various algorithm parameters given to the structure CvStarDetectorParams:}
236 typedef struct CvStarDetectorParams
238 int maxSize; // maximal size of the features detected. The following
239 // values of the parameter are supported:
240 // 4, 6, 8, 11, 12, 16, 22, 23, 32, 45, 46, 64, 90, 128
241 int responseThreshold; // threshold for the approximatd laplacian,
242 // used to eliminate weak features
243 int lineThresholdProjected; // another threshold for laplacian to
245 int lineThresholdBinarized; // another threshold for the feature
246 // scale to eliminate edges
247 int suppressNonmaxSize; // linear size of a pixel neighborhood
248 // for non-maxima suppression
250 CvStarDetectorParams;
254 The function GetStarKeypoints extracts keypoints that are local
255 scale-space extremas. The scale-space is constructed by computing
256 approximate values of laplacians with different sigma's at each
257 pixel. Instead of using pyramids, a popular approach to save computing
258 time, all of the laplacians are computed at each pixel of the original
259 high-resolution image. But each approximate laplacian value is computed
260 in O(1) time regardless of the sigma, thanks to the use of integral
261 images. The algorithm is based on the paper
264 of a square, hexagon or octagon it uses an 8-end star shape, hence the name,
265 consisting of overlapping upright and tilted squares.
267 Each computed feature is represented by the following structure:
270 typedef struct CvStarKeypoint
272 CvPoint pt; // coordinates of the feature
273 int size; // feature size, see CvStarDetectorParams::maxSize
274 float response; // the approximated laplacian value at that point.
278 inline CvStarKeypoint cvStarKeypoint(CvPoint pt, int size, float response);
282 Below is the small usage sample:
288 int main(int argc, char** argv)
290 const char* filename = argc > 1 ? argv[1] : "lena.jpg";
291 IplImage* img = cvLoadImage( filename, 0 ), *cimg;
292 CvMemStorage* storage = cvCreateMemStorage(0);
293 CvSeq* keypoints = 0;
298 cvNamedWindow( "image", 1 );
299 cvShowImage( "image", img );
300 cvNamedWindow( "features", 1 );
301 cimg = cvCreateImage( cvGetSize(img), 8, 3 );
302 cvCvtColor( img, cimg, CV_GRAY2BGR );
304 keypoints = cvGetStarKeypoints( img, storage, cvStarDetectorParams(45) );
306 for( i = 0; i < (keypoints ? keypoints->total : 0); i++ )
308 CvStarKeypoint kpt = *(CvStarKeypoint*)cvGetSeqElem(keypoints, i);
310 cvCircle( cimg, kpt.pt, r, CV_RGB(0,255,0));
311 cvLine( cimg, cvPoint(kpt.pt.x + r, kpt.pt.y + r),
312 cvPoint(kpt.pt.x - r, kpt.pt.y - r), CV_RGB(0,255,0));
313 cvLine( cimg, cvPoint(kpt.pt.x - r, kpt.pt.y + r),
314 cvPoint(kpt.pt.x + r, kpt.pt.y - r), CV_RGB(0,255,0));
316 cvShowImage( "features", cimg );
322 \cvCPyFunc{GoodFeaturesToTrack}
323 Determines strong corners on an image.
326 void cvGoodFeaturesToTrack(
327 \par const CvArr* image
328 \par CvArr* eigImage, CvArr* tempImage
329 \par CvPoint2D32f* corners
330 \par int* cornerCount
331 \par double qualityLevel
332 \par double minDistance
333 \par const CvArr* mask=NULL
336 \par double k=0.04 );
338 \cvdefPy{GoodFeaturesToTrack(image,eigImage,tempImage,cornerCount,qualityLevel,minDistance,mask=NULL,blockSize=3,useHarris=0,k=0.04)-> corners}
341 \cvarg{image}{The source 8-bit or floating-point 32-bit, single-channel image}
342 \cvarg{eigImage}{Temporary floating-point 32-bit image, the same size as \texttt{image}}
343 \cvarg{tempImage}{Another temporary image, the same size and format as \texttt{eigImage}}
345 \cvarg{corners}{Output parameter; detected corners}
346 \cvarg{cornerCount}{Output parameter; number of detected corners}
348 \cvarg{cornerCount}{number of corners to detect}
350 \cvarg{qualityLevel}{Multiplier for the max/min eigenvalue; specifies the minimal accepted quality of image corners}
351 \cvarg{minDistance}{Limit, specifying the minimum possible distance between the returned corners; Euclidian distance is used}
352 \cvarg{mask}{Region of interest. The function selects points either in the specified region or in the whole image if the mask is NULL}
353 \cvarg{blockSize}{Size of the averaging block, passed to the underlying \cvCPyCross{CornerMinEigenVal} or \cvCPyCross{CornerHarris} used by the function}
354 \cvarg{useHarris}{If nonzero, Harris operator (\cvCPyCross{CornerHarris}) is used instead of default \cvCPyCross{CornerMinEigenVal}}
355 \cvarg{k}{Free parameter of Harris detector; used only if ($\texttt{useHarris} != 0$)}
358 The function finds the corners with big eigenvalues in the image. The function first calculates the minimal
359 eigenvalue for every source image pixel using the \cvCPyCross{CornerMinEigenVal}
360 function and stores them in \texttt{eigImage}. Then it performs
361 non-maxima suppression (only the local maxima in $3\times 3$ neighborhood
362 are retained). The next step rejects the corners with the minimal
363 eigenvalue less than $\texttt{qualityLevel} \cdot max(\texttt{eigImage}(x,y))$.
364 Finally, the function ensures that the distance between any two corners is not smaller than \texttt{minDistance}. The weaker corners (with a smaller min eigenvalue) that are too close to the stronger corners are rejected.
366 Note that the if the function is called with different values \texttt{A} and \texttt{B} of the parameter \texttt{qualityLevel}, and \texttt{A} > {B}, the array of returned corners with \texttt{qualityLevel=A} will be the prefix of the output corners array with \texttt{qualityLevel=B}.
368 \cvCPyFunc{HoughLines2}
369 Finds lines in a binary image using a Hough transform.
372 CvSeq* cvHoughLines2( \par CvArr* image,\par void* storage,\par int method,\par double rho,\par double theta,\par int threshold,\par double param1=0,\par double param2=0 );
374 \cvdefPy{HoughLines2(image,storage,method,rho,theta,threshold,param1=0,param2=0)-> lines}
377 \cvarg{image}{The 8-bit, single-channel, binary source image. In the case of a probabilistic method, the image is modified by the function}
378 \cvarg{storage}{The storage for the lines that are detected. It can
379 be a memory storage (in this case a sequence of lines is created in
380 the storage and returned by the function) or single row/single column
381 matrix (CvMat*) of a particular type (see below) to which the lines'
382 parameters are written. The matrix header is modified by the function
383 so its \texttt{cols} or \texttt{rows} will contain the number of lines
384 detected. If \texttt{storage} is a matrix and the actual number
385 of lines exceeds the matrix size, the maximum possible number of lines
386 is returned (in the case of standard hough transform the lines are sorted
387 by the accumulator value)}
388 \cvarg{method}{The Hough transform variant, one of the following:
390 \cvarg{CV\_HOUGH\_STANDARD}{classical or standard Hough transform. Every line is represented by two floating-point numbers $(\rho, \theta)$, where $\rho$ is a distance between (0,0) point and the line, and $\theta$ is the angle between x-axis and the normal to the line. Thus, the matrix must be (the created sequence will be) of \texttt{CV\_32FC2} type}
391 \cvarg{CV\_HOUGH\_PROBABILISTIC}{probabilistic Hough transform (more efficient in case if picture contains a few long linear segments). It returns line segments rather than the whole line. Each segment is represented by starting and ending points, and the matrix must be (the created sequence will be) of \texttt{CV\_32SC4} type}
392 \cvarg{CV\_HOUGH\_MULTI\_SCALE}{multi-scale variant of the classical Hough transform. The lines are encoded the same way as \texttt{CV\_HOUGH\_STANDARD}}
394 \cvarg{rho}{Distance resolution in pixel-related units}
395 \cvarg{theta}{Angle resolution measured in radians}
396 \cvarg{threshold}{Threshold parameter. A line is returned by the function if the corresponding accumulator value is greater than \texttt{threshold}}
397 \cvarg{param1}{The first method-dependent parameter:
399 \item For the classical Hough transform it is not used (0).
400 \item For the probabilistic Hough transform it is the minimum line length.
401 \item For the multi-scale Hough transform it is the divisor for the distance resolution $\rho$. (The coarse distance resolution will be $\rho$ and the accurate resolution will be $(\rho / \texttt{param1})$).
403 \cvarg{param2}{The second method-dependent parameter:
405 \item For the classical Hough transform it is not used (0).
406 \item For the probabilistic Hough transform it is the maximum gap between line segments lying on the same line to treat them as a single line segment (i.e. to join them).
407 \item For the multi-scale Hough transform it is the divisor for the angle resolution $\theta$. (The coarse angle resolution will be $\theta$ and the accurate resolution will be $(\theta / \texttt{param2})$).
411 The function implements a few variants of the Hough transform for line detection.
414 \textbf{Example. Detecting lines with Hough transform.}
416 /* This is a standalone program. Pass an image name as a first parameter
417 of the program. Switch between standard and probabilistic Hough transform
418 by changing "#if 1" to "#if 0" and back */
423 int main(int argc, char** argv)
426 if( argc == 2 && (src=cvLoadImage(argv[1], 0))!= 0)
428 IplImage* dst = cvCreateImage( cvGetSize(src), 8, 1 );
429 IplImage* color_dst = cvCreateImage( cvGetSize(src), 8, 3 );
430 CvMemStorage* storage = cvCreateMemStorage(0);
433 cvCanny( src, dst, 50, 200, 3 );
434 cvCvtColor( dst, color_dst, CV_GRAY2BGR );
436 lines = cvHoughLines2( dst,
445 for( i = 0; i < MIN(lines->total,100); i++ )
447 float* line = (float*)cvGetSeqElem(lines,i);
449 float theta = line[1];
451 double a = cos(theta), b = sin(theta);
452 double x0 = a*rho, y0 = b*rho;
453 pt1.x = cvRound(x0 + 1000*(-b));
454 pt1.y = cvRound(y0 + 1000*(a));
455 pt2.x = cvRound(x0 - 1000*(-b));
456 pt2.y = cvRound(y0 - 1000*(a));
457 cvLine( color_dst, pt1, pt2, CV_RGB(255,0,0), 3, 8 );
460 lines = cvHoughLines2( dst,
462 CV_HOUGH_PROBABILISTIC,
468 for( i = 0; i < lines->total; i++ )
470 CvPoint* line = (CvPoint*)cvGetSeqElem(lines,i);
471 cvLine( color_dst, line[0], line[1], CV_RGB(255,0,0), 3, 8 );
474 cvNamedWindow( "Source", 1 );
475 cvShowImage( "Source", src );
477 cvNamedWindow( "Hough", 1 );
478 cvShowImage( "Hough", color_dst );
485 This is the sample picture the function parameters have been tuned for:
487 \includegraphics[width=0.5\textwidth]{pics/building.jpg}
489 And this is the output of the above program in the case of probabilistic Hough transform (\texttt{\#if 0} case):
491 \includegraphics[width=0.5\textwidth]{pics/houghp.png}
494 \cvCPyFunc{PreCornerDetect}
495 Calculates the feature map for corner detection.
498 void cvPreCornerDetect(
499 \par const CvArr* image,
501 \par int apertureSize=3 );
503 \cvdefPy{PreCornerDetect(image,corners,apertureSize=3)-> None}
505 \cvarg{image}{Input image}
506 \cvarg{corners}{Image to store the corner candidates}
507 \cvarg{apertureSize}{Aperture parameter for the Sobel operator (see \cvCPyCross{Sobel})}
510 The function calculates the function
513 D_x^2 D_{yy} + D_y^2 D_{xx} - 2 D_x D_y D_{xy}
516 where $D_?$ denotes one of the first image derivatives and $D_{??}$ denotes a second image derivative.
518 The corners can be found as local maximums of the function below:
521 // assume that the image is floating-point
522 IplImage* corners = cvCloneImage(image);
523 IplImage* dilated_corners = cvCloneImage(image);
524 IplImage* corner_mask = cvCreateImage( cvGetSize(image), 8, 1 );
525 cvPreCornerDetect( image, corners, 3 );
526 cvDilate( corners, dilated_corners, 0, 1 );
527 cvSubS( corners, dilated_corners, corners );
528 cvCmpS( corners, 0, corner_mask, CV_CMP_GE );
529 cvReleaseImage( &corners );
530 cvReleaseImage( &dilated_corners );
534 \cvCPyFunc{SampleLine}
535 Reads the raster line to the buffer.
539 \par const CvArr* image
543 \par int connectivity=8 );
547 \cvarg{image}{Image to sample the line from}
548 \cvarg{pt1}{Starting line point}
549 \cvarg{pt2}{Ending line point}
550 \cvarg{buffer}{Buffer to store the line points; must have enough size to store
551 $max( |\texttt{pt2.x} - \texttt{pt1.x}|+1, |\texttt{pt2.y} - \texttt{pt1.y}|+1 )$
552 points in the case of an 8-connected line and
553 $ (|\texttt{pt2.x}-\texttt{pt1.x}|+|\texttt{pt2.y}-\texttt{pt1.y}|+1) $
554 in the case of a 4-connected line}
555 \cvarg{connectivity}{The line connectivity, 4 or 8}
558 The function implements a particular application of line iterators. The function reads all of the image points lying on the line between \texttt{pt1} and \texttt{pt2}, including the end points, and stores them into the buffer.
568 Finds edges in an image using Canny algorithm.
570 \cvdefCpp{void Canny( const Mat\& image, Mat\& edges,\par
571 double threshold1, double threshold2,\par
572 int apertureSize=3, bool L2gradient=false );}
574 \cvarg{image}{Single-channel 8-bit input image}
575 \cvarg{edges}{The output edge map. It will have the same size and the same type as \texttt{image}}
576 \cvarg{threshold1}{The first threshold for the hysteresis procedure}
577 \cvarg{threshold2}{The second threshold for the hysteresis procedure}
578 \cvarg{apertureSize}{Aperture size for the \cvCppCross{Sobel} operator}
579 \cvarg{L2gradient}{Indicates, whether the more accurate $L_2$ norm $=\sqrt{(dI/dx)^2 + (dI/dy)^2}$ should be used to compute the image gradient magnitude (\texttt{L2gradient=true}), or a faster default $L_1$ norm $=|dI/dx|+|dI/dy|$ is enough (\texttt{L2gradient=false})}
582 The function finds edges in the input image \texttt{image} and marks them in the output map \texttt{edges} using the Canny algorithm. The smallest value between \texttt{threshold1} and \texttt{threshold2} is used for edge linking, the largest value is used to find the initial segments of strong edges, see
583 \url{http://en.wikipedia.org/wiki/Canny_edge_detector}
585 \cvCppFunc{cornerEigenValsAndVecs}
586 Calculates eigenvalues and eigenvectors of image blocks for corner detection.
588 \cvdefCpp{void cornerEigenValsAndVecs( const Mat\& src, Mat\& dst,\par
589 int blockSize, int apertureSize,\par
590 int borderType=BORDER\_DEFAULT );}
592 \cvarg{src}{Input single-channel 8-bit or floating-point image}
593 \cvarg{dst}{Image to store the results. It will have the same size as \texttt{src} and the type \texttt{CV\_32FC(6)}}
594 \cvarg{blockSize}{Neighborhood size (see discussion)}
595 \cvarg{apertureSize}{Aperture parameter for the \cvCppCross{Sobel} operator}
596 \cvarg{boderType}{Pixel extrapolation method; see \cvCppCross{borderInterpolate}}
599 For every pixel $p$, the function \texttt{cornerEigenValsAndVecs} considers a \texttt{blockSize} $\times$ \texttt{blockSize} neigborhood $S(p)$. It calculates the covariation matrix of derivatives over the neighborhood as:
603 \sum_{S(p)}(dI/dx)^2 & \sum_{S(p)}(dI/dx dI/dy)^2 \\
604 \sum_{S(p)}(dI/dx dI/dy)^2 & \sum_{S(p)}(dI/dy)^2
608 Where the derivatives are computed using \cvCppCross{Sobel} operator.
610 After that it finds eigenvectors and eigenvalues of $M$ and stores them into destination image in the form
611 $(\lambda_1, \lambda_2, x_1, y_1, x_2, y_2)$ where
613 \item[$\lambda_1, \lambda_2$]are the eigenvalues of $M$; not sorted
614 \item[$x_1, y_1$]are the eigenvectors corresponding to $\lambda_1$
615 \item[$x_2, y_2$]are the eigenvectors corresponding to $\lambda_2$
618 The output of the function can be used for robust edge or corner detection.
620 See also: \cvCppCross{cornerMinEigenVal}, \cvCppCross{cornerHarris}, \cvCppCross{preCornerDetect}
622 \cvCppFunc{cornerHarris}
623 Harris edge detector.
625 \cvdefCpp{void cornerHarris( const Mat\& src, Mat\& dst, int blockSize,\par
626 int apertureSize, double k,\par
627 int borderType=BORDER\_DEFAULT );}
629 \cvarg{src}{Input single-channel 8-bit or floating-point image}
630 \cvarg{dst}{Image to store the Harris detector responses; will have type \texttt{CV\_32FC1} and the same size as \texttt{src}}
631 \cvarg{blockSize}{Neighborhood size (see the discussion of \cvCppCross{cornerEigenValsAndVecs})}
632 \cvarg{apertureSize}{Aperture parameter for the \cvCppCross{Sobel} operator}
633 \cvarg{k}{Harris detector free parameter. See the formula below}
634 \cvarg{boderType}{Pixel extrapolation method; see \cvCppCross{borderInterpolate}}
637 The function runs the Harris edge detector on the image. Similarly to \cvCppCross{cornerMinEigenVal} and \cvCppCross{cornerEigenValsAndVecs}, for each pixel $(x, y)$ it calculates a $2\times2$ gradient covariation matrix $M^{(x,y)}$ over a $\texttt{blockSize} \times \texttt{blockSize}$ neighborhood. Then, it computes the following characteristic:
640 \texttt{dst}(x,y) = \mathrm{det} M^{(x,y)} - k \cdot \left(\mathrm{tr} M^{(x,y)}\right)^2
643 Corners in the image can be found as the local maxima of this response map.
645 \cvCppFunc{cornerMinEigenVal}
646 Calculates the minimal eigenvalue of gradient matrices for corner detection.
648 \cvdefCpp{void cornerMinEigenVal( const Mat\& src, Mat\& dst,\par
649 int blockSize, int apertureSize=3,\par
650 int borderType=BORDER\_DEFAULT );}
652 \cvarg{src}{Input single-channel 8-bit or floating-point image}
653 \cvarg{dst}{Image to store the minimal eigenvalues; will have type \texttt{CV\_32FC1} and the same size as \texttt{src}}
654 \cvarg{blockSize}{Neighborhood size (see the discussion of \cvCppCross{cornerEigenValsAndVecs})}
655 \cvarg{apertureSize}{Aperture parameter for the \cvCppCross{Sobel} operator}
656 \cvarg{boderType}{Pixel extrapolation method; see \cvCppCross{borderInterpolate}}
659 The function is similar to \cvCppCross{cornerEigenValsAndVecs} but it calculates and stores only the minimal eigenvalue of the covariation matrix of derivatives, i.e. $\min(\lambda_1, \lambda_2)$ in terms of the formulae in \cvCppCross{cornerEigenValsAndVecs} description.
661 \cvCppFunc{cornerSubPix}
662 Refines the corner locations.
664 \cvdefCpp{void cornerSubPix( const Mat\& image, vector<Point2f>\& corners,\par
665 Size winSize, Size zeroZone,\par
666 TermCriteria criteria );}
668 \cvarg{image}{Input image}
669 \cvarg{corners}{Initial coordinates of the input corners; refined coordinates on output}
670 \cvarg{winSize}{Half of the side length of the search window. For example, if \texttt{winSize=Size(5,5)}, then a $5*2+1 \times 5*2+1 = 11 \times 11$ search window would be used}
671 \cvarg{zeroZone}{Half of the size of the dead region in the middle of the search zone over which the summation in the formula below is not done. It is used sometimes to avoid possible singularities of the autocorrelation matrix. The value of (-1,-1) indicates that there is no such size}
672 \cvarg{criteria}{Criteria for termination of the iterative process of corner refinement. That is, the process of corner position refinement stops either after a certain number of iterations or when a required accuracy is achieved. The \texttt{criteria} may specify either of or both the maximum number of iteration and the required accuracy}
675 The function iterates to find the sub-pixel accurate location of corners, or radial saddle points, as shown in on the picture below.
677 \includegraphics[width=1.0\textwidth]{pics/cornersubpix.png}
679 Sub-pixel accurate corner locator is based on the observation that every vector from the center $q$ to a point $p$ located within a neighborhood of $q$ is orthogonal to the image gradient at $p$ subject to image and measurement noise. Consider the expression:
682 \epsilon_i = {DI_{p_i}}^T \cdot (q - p_i)
685 where ${DI_{p_i}}$ is the image gradient at the one of the points $p_i$ in a neighborhood of $q$. The value of $q$ is to be found such that $\epsilon_i$ is minimized. A system of equations may be set up with $\epsilon_i$ set to zero:
688 \sum_i(DI_{p_i} \cdot {DI_{p_i}}^T) - \sum_i(DI_{p_i} \cdot {DI_{p_i}}^T \cdot p_i)
691 where the gradients are summed within a neighborhood ("search window") of $q$. Calling the first gradient term $G$ and the second gradient term $b$ gives:
697 The algorithm sets the center of the neighborhood window at this new center $q$ and then iterates until the center keeps within a set threshold.
700 \cvCppFunc{goodFeaturesToTrack}
701 Determines strong corners on an image.
703 \cvdefCpp{void goodFeaturesToTrack( const Mat\& image, vector<Point2f>\& corners,\par
704 int maxCorners, double qualityLevel, double minDistance,\par
705 const Mat\& mask=Mat(), int blockSize=3,\par
706 bool useHarrisDetector=false, double k=0.04 );}
708 \cvarg{image}{The input 8-bit or floating-point 32-bit, single-channel image}
709 \cvarg{corners}{The output vector of detected corners}
710 \cvarg{maxCorners}{The maximum number of corners to return. If there are more corners than that will be found, the strongest of them will be returned}
711 \cvarg{qualityLevel}{Characterizes the minimal accepted quality of image corners; the value of the parameter is multiplied by the by the best corner quality measure (which is the min eigenvalue, see \cvCppCross{cornerMinEigenVal}, or the Harris function response, see \cvCppCross{cornerHarris}). The corners, which quality measure is less than the product, will be rejected. For example, if the best corner has the quality measure = 1500, and the \texttt{qualityLevel=0.01}, then all the corners which quality measure is less than 15 will be rejected.}
712 \cvarg{minDistance}{The minimum possible Euclidean distance between the returned corners}
713 \cvarg{mask}{The optional region of interest. If the image is not empty (then it needs to have the type \texttt{CV\_8UC1} and the same size as \texttt{image}), it will specify the region in which the corners are detected}
714 \cvarg{blockSize}{Size of the averaging block for computing derivative covariation matrix over each pixel neighborhood, see \cvCppCross{cornerEigenValsAndVecs}}
715 \cvarg{useHarrisDetector}{Indicates, whether to use \hyperref[cornerHarris]{Harris} operator or \cvCppCross{cornerMinEigenVal}}
716 \cvarg{k}{Free parameter of Harris detector}
719 The function finds the most prominent corners in the image or in the specified image region, as described
722 \item the function first calculates the corner quality measure at every source image pixel using the \cvCppCross{cornerMinEigenVal} or \cvCppCross{cornerHarris}
723 \item then it performs non-maxima suppression (the local maxima in $3\times 3$ neighborhood
725 \item the next step rejects the corners with the minimal eigenvalue less than $\texttt{qualityLevel} \cdot \max_{x,y} qualityMeasureMap(x,y)$.
726 \item the remaining corners are then sorted by the quality measure in the descending order.
727 \item finally, the function throws away each corner $pt_j$ if there is a stronger corner $pt_i$ ($i < j$) such that the distance between them is less than \texttt{minDistance}
730 The function can be used to initialize a point-based tracker of an object.
732 Note that the if the function is called with different values \texttt{A} and \texttt{B} of the parameter \texttt{qualityLevel}, and \texttt{A} > {B}, the vector of returned corners with \texttt{qualityLevel=A} will be the prefix of the output vector with \texttt{qualityLevel=B}.
734 See also: \cvCppCross{cornerMinEigenVal}, \cvCppCross{cornerHarris}, \cvCppCross{calcOpticalFlowPyrLK}, \cvCppCross{estimateRigidMotion}, \cvCppCross{PlanarObjectDetector}, \cvCppCross{OneWayDescriptor}
736 \cvCppFunc{HoughCircles}
737 Finds circles in a grayscale image using a Hough transform.
739 \cvdefCpp{void HoughCircles( Mat\& image, vector<Vec3f>\& circles,\par
740 int method, double dp, double minDist,\par
741 double param1=100, double param2=100,\par
742 int minRadius=0, int maxRadius=0 );}
744 \cvarg{image}{The 8-bit, single-channel, grayscale input image}
745 \cvarg{circles}{The output vector of found circles. Each vector is encoded as 3-element floating-point vector $(x, y, radius)$}
746 \cvarg{method}{Currently, the only implemented method is \texttt{CV\_HOUGH\_GRADIENT}, which is basically \emph{21HT}, described in \cite{Yuen90}.}
747 \cvarg{dp}{The inverse ratio of the accumulator resolution to the image resolution. For example, if \texttt{dp=1}, the accumulator will have the same resolution as the input image, if \texttt{dp=2} - accumulator will have half as big width and height, etc}
748 \cvarg{minDist}{Minimum distance between the centers of the detected circles. If the parameter is too small, multiple neighbor circles may be falsely detected in addition to a true one. If it is too large, some circles may be missed}
749 \cvarg{param1}{The first method-specific parameter. in the case of \texttt{CV\_HOUGH\_GRADIENT} it is the higher threshold of the two passed to \cvCppCross{Canny} edge detector (the lower one will be twice smaller)}
750 \cvarg{param2}{The second method-specific parameter. in the case of \texttt{CV\_HOUGH\_GRADIENT} it is the accumulator threshold at the center detection stage. The smaller it is, the more false circles may be detected. Circles, corresponding to the larger accumulator values, will be returned first}
751 \cvarg{minRadius}{Minimum circle radius}
752 \cvarg{maxRadius}{Maximum circle radius}
755 The function finds circles in a grayscale image using some modification of Hough transform. Here is a short usage example:
764 int main(int argc, char** argv)
767 if( argc != 2 && !(img=imread(argv[1], 1)).data)
769 cvtColor(img, gray, CV_BGR2GRAY);
770 // smooth it, otherwise a lot of false circles may be detected
771 GaussianBlur( gray, gray, 9, 9, 2, 2 );
772 vector<Vec3f> circles;
773 houghCircles(gray, circles, CV_HOUGH_GRADIENT,
774 2, gray->rows/4, 200, 100 );
775 for( size_t i = 0; i < circles.size(); i++ )
777 Point center(cvRound(circles[i][0]), cvRound(circles[i][1]));
778 int radius = cvRound(circles[i][2]);
779 // draw the circle center
780 circle( img, center, 3, Scalar(0,255,0), -1, 8, 0 );
781 // draw the circle outline
782 circle( img, center, radius, Scalar(0,0,255), 3, 8, 0 );
784 namedWindow( "circles", 1 );
785 imshow( "circles", img );
790 Note that usually the function detects the circles' centers well, however it may fail to find the correct radii. You can assist the function by specifying the radius range (\texttt{minRadius} and \texttt{maxRadius}) if you know it, or you may ignore the returned radius, use only the center and find the correct radius using some additional procedure.
792 See also: \cvCppCross{fitEllipse}, \cvCppCross{minEnclosingCircle}
794 \cvCppFunc{HoughLines}
795 Finds lines in a binary image using standard Hough transform.
797 \cvdefCpp{void HoughLines( Mat\& image, vector<Vec2f>\& lines,\par
798 double rho, double theta, int threshold,\par
799 double srn=0, double stn=0 );}
801 \cvarg{image}{The 8-bit, single-channel, binary source image. The image may be modified by the function}
802 \cvarg{lines}{The output vector of lines. Each line is represented by a two-element vector $(\rho, \theta)$. $\rho$ is the distance from the coordinate origin $(0,0)$ (top-left corner of the image) and $\theta$ is the line rotation angle in radians ($0 \sim \textrm{vertical line}, \pi/2 \sim \textrm{horizontal line}$)}
803 \cvarg{rho}{Distance resolution of the accumulator in pixels}
804 \cvarg{theta}{Angle resolution of the accumulator in radians}
805 \cvarg{threshold}{The accumulator threshold parameter. Only those lines are returned that get enough votes ($>\texttt{threshold}$)}
806 \cvarg{srn}{For the multi-scale Hough transform it is the divisor for the distance resolution \texttt{rho}. The coarse accumulator distance resolution will be \texttt{rho} and the accurate accumulator resolution will be \texttt{rho/srn}. If both \texttt{srn=0} and \texttt{stn=0} then the classical Hough transform is used, otherwise both these parameters should be positive.}
807 \cvarg{stn}{For the multi-scale Hough transform it is the divisor for the distance resolution \texttt{theta}}
810 The function implements standard or standard multi-scale Hough transform algorithm for line detection. See \cvCppCross{HoughLinesP} for the code example.
813 \cvCppFunc{HoughLinesP}
814 Finds lines segments in a binary image using probabilistic Hough transform.
816 \cvdefCpp{void HoughLinesP( Mat\& image, vector<Vec4i>\& lines,\par
817 double rho, double theta, int threshold,\par
818 double minLineLength=0, double maxLineGap=0 );}
820 \cvarg{image}{The 8-bit, single-channel, binary source image. The image may be modified by the function}
821 \cvarg{lines}{The output vector of lines. Each line is represented by a 4-element vector $(x_1, y_1, x_2, y_2)$, where $(x_1,y_1)$ and $(x_2, y_2)$ are the ending points of each line segment detected.}
822 \cvarg{rho}{Distance resolution of the accumulator in pixels}
823 \cvarg{theta}{Angle resolution of the accumulator in radians}
824 \cvarg{threshold}{The accumulator threshold parameter. Only those lines are returned that get enough votes ($>\texttt{threshold}$)}
825 \cvarg{minLineLength}{The minimum line length. Line segments shorter than that will be rejected}
826 \cvarg{maxLineGap}{The maximum allowed gap between points on the same line to link them.}
829 The function implements probabilistic Hough transform algorithm for line detection, described in \cite{Matas00}. Below is line detection example:
832 /* This is a standalone program. Pass an image name as a first parameter
833 of the program. Switch between standard and probabilistic Hough transform
834 by changing "#if 1" to "#if 0" and back */
841 int main(int argc, char** argv)
843 Mat src, dst, color_dst;
844 if( argc != 2 || !(src=imread(argv[1], 0)).data)
847 Canny( src, dst, 50, 200, 3 );
848 cvtColor( dst, color_dst, CV_GRAY2BGR );
852 HoughLines( dst, lines, 1, CV_PI/180, 100 );
854 for( size_t i = 0; i < lines.size(); i++ )
856 float rho = lines[i][0];
857 float theta = lines[i][1];
858 double a = cos(theta), b = sin(theta);
859 double x0 = a*rho, y0 = b*rho;
860 Point pt1(cvRound(x0 + 1000*(-b)),
861 cvRound(y0 + 1000*(a)));
862 Point pt2(cvRound(x0 - 1000*(-b)),
863 cvRound(y0 - 1000*(a)));
864 line( color_dst, pt1, pt2, Scalar(0,0,255), 3, 8 );
868 HoughLinesP( dst, lines, 1, CV_PI/180, 80, 30, 10 );
869 for( size_t i = 0; i < lines.size(); i++ )
871 line( color_dst, Point(lines[i][0], lines[i][1]),
872 Point(lines[i][2], lines[i][3]), Scalar(0,0,255), 3, 8 );
875 namedWindow( "Source", 1 );
876 imshow( "Source", src );
878 namedWindow( "Detected Lines", 1 );
879 imshow( "Detected Lines", color_dst );
887 This is the sample picture the function parameters have been tuned for:
889 \includegraphics[width=0.5\textwidth]{pics/building.jpg}
891 And this is the output of the above program in the case of probabilistic Hough transform
893 \includegraphics[width=0.5\textwidth]{pics/houghp.png}
895 \cvCppFunc{perCornerDetect}
896 Calculates the feature map for corner detection
898 \cvdefCpp{void preCornerDetect( const Mat\& src, Mat\& dst, int apertureSize,\par
899 int borderType=BORDER\_DEFAULT );}
901 \cvarg{src}{The source single-channel 8-bit of floating-point image}
902 \cvarg{dst}{The output image; will have type \texttt{CV\_32F} and the same size as \texttt{src}}
903 \cvarg{apertureSize}{Aperture size of \cvCppCross{Sobel}}
904 \cvarg{borderType}{The pixel extrapolation method; see \cvCppCross{borderInterpolate}}
907 The function calculates the complex spatial derivative-based function of the source image
910 \texttt{dst} = (D_x \texttt{src})^2 \cdot D_{yy} \texttt{src} + (D_y \texttt{src})^2 \cdot D_{xx} \texttt{src} - 2 D_x \texttt{src} \cdot D_y \texttt{src} \cdot D_{xy} \texttt{src}
913 where $D_x$, $D_y$ are the first image derivatives, $D_{xx}$, $D_{yy}$ are the second image derivatives and $D_{xy}$ is the mixed derivative.
915 The corners can be found as local maximums of the functions, as shown below:
918 Mat corners, dilated_corners;
919 preCornerDetect(image, corners, 3);
920 // dilation with 3x3 rectangular structuring element
921 dilate(corners, dilated_corners, Mat(), 1);
922 Mat corner_mask = corners == dilated_corners;
927 Data structure for salient point detectors
933 // default constructor
935 // two complete constructors
936 KeyPoint(Point2f _pt, float _size, float _angle=-1,
937 float _response=0, int _octave=0, int _class_id=-1);
938 KeyPoint(float x, float y, float _size, float _angle=-1,
939 float _response=0, int _octave=0, int _class_id=-1);
940 // coordinate of the point
944 // feature orintation in degrees
945 // (has negative value if the orientation
946 // is not defined/not computed)
949 // (can be used to select only
950 // the most prominent key points)
952 // scale-space octave in which the feature has been found;
953 // may correlate with the size
955 // point (can be used by feature
956 // classifiers or object detectors)
960 // reading/writing a vector of keypoints to a file storage
961 void write(FileStorage& fs, const string& name, const vector<KeyPoint>& keypoints);
962 void read(const FileNode& node, vector<KeyPoint>& keypoints);
967 Maximally-Stable Extremal Region Extractor
970 class MSER : public CvMSERParams
973 // default constructor
975 // constructor that initializes all the algorithm parameters
976 MSER( int _delta, int _min_area, int _max_area,
977 float _max_variation, float _min_diversity,
978 int _max_evolution, double _area_threshold,
979 double _min_margin, int _edge_blur_size );
980 // runs the extractor on the specified image; returns the MSERs,
981 // each encoded as a contour (vector<Point>, see findContours)
982 // the optional mask marks the area where MSERs are searched for
983 void operator()(Mat& image, vector<vector<Point> >& msers, const Mat& mask) const;
987 The class encapsulates all the parameters of MSER (see \url{http://en.wikipedia.org/wiki/Maximally_stable_extremal_regions}) extraction algorithm.
990 Class for extracting Speeded Up Robust Features from an image.
993 class SURF : public CvSURFParams
996 // default constructor
998 // constructor that initializes all the algorithm parameters
999 SURF(double _hessianThreshold, int _nOctaves=4,
1000 int _nOctaveLayers=2, bool _extended=false);
1001 // returns the number of elements in each descriptor (64 or 128)
1002 int descriptorSize() const;
1003 // detects keypoints using fast multi-scale Hessian detector
1004 void operator()(const Mat& img, const Mat& mask,
1005 vector<KeyPoint>& keypoints) const;
1006 // detects keypoints and computes the SURF descriptors for them
1007 void operator()(const Mat& img, const Mat& mask,
1008 vector<KeyPoint>& keypoints,
1009 vector<float>& descriptors,
1010 bool useProvidedKeypoints=false) const;
1014 The class \texttt{SURF} implements Speeded Up Robust Features descriptor \cite{Bay06}.
1015 There is fast multi-scale Hessian keypoint detector that can be used to find the keypoints
1016 (which is the default option), but the descriptors can be also computed for the user-specified keypoints.
1017 The function can be used for object tracking and localization, image stitching etc. See the
1018 \texttt{find\_obj.cpp} demo in OpenCV samples directory.
1021 \cvclass{StarDetector}
1022 Implements Star keypoint detector
1025 class StarDetector : CvStarDetectorParams
1028 // default constructor
1030 // the full constructor initialized all the algorithm parameters:
1031 // maxSize - maximum size of the features. The following
1032 // values of the parameter are supported:
1033 // 4, 6, 8, 11, 12, 16, 22, 23, 32, 45, 46, 64, 90, 128
1034 // responseThreshold - threshold for the approximated laplacian,
1035 // used to eliminate weak features. The larger it is,
1036 // the less features will be retrieved
1037 // lineThresholdProjected - another threshold for the laplacian to
1039 // lineThresholdBinarized - another threshold for the feature
1040 // size to eliminate edges.
1041 // The larger the 2 threshold, the more points you get.
1042 StarDetector(int maxSize, int responseThreshold,
1043 int lineThresholdProjected,
1044 int lineThresholdBinarized,
1045 int suppressNonmaxSize);
1047 // finds keypoints in an image
1048 void operator()(const Mat& image, vector<KeyPoint>& keypoints) const;
1052 The class implements a modified version of CenSurE keypoint detector described in