Skip to content

O
opencv_contrib
Commit df859520 authored by Muresan Mircea Paul's avatar Muresan Mircea Paul
Browse files
Options

made modifications in accordance to the comments 

fixed that unused variable warning

fixed windows warnings

Added 2 samples just to show how to access functionality

fixed issues

added the up to date version of sbm

modified samples to be warning free
parent 54d41251
Hide whitespace changes
Inline Side-by-side
Showing with 528 additions and 573 deletions
modules/stereo/include/opencv2/stereo.hpp
View file @ df859520
......@@ -47,211 +47,229 @@
#include "opencv2/core.hpp"
#include "opencv2/features2d.hpp"
#include "opencv2/core/affine.hpp"
#include "../../stereo/src/descriptor.hpp"
#include "../../stereo/src/matching.hpp"
/**
@defgroup stereo Stereo Correspondance Algorithms
*/
*/
namespace cv
{
namespace stereo
{
//! @addtogroup stereo
//! @{
// void correctMatches( InputArray F, InputArray points1, InputArray points2,
// OutputArray newPoints1, OutputArray newPoints2 );
/** @brief Filters off small noise blobs (speckles) in the disparity map
@param img The input 16-bit signed disparity image
@param newVal The disparity value used to paint-off the speckles
@param maxSpeckleSize The maximum speckle size to consider it a speckle. Larger blobs are not
affected by the algorithm
@param maxDiff Maximum difference between neighbor disparity pixels to put them into the same
blob. Note that since StereoBM, StereoSGBM and may be other algorithms return a fixed-point
disparity map, where disparity values are multiplied by 16, this scale factor should be taken into
account when specifying this parameter value.
@param buf The optional temporary buffer to avoid memory allocation within the function.
*/
/** @brief The base class for stereo correspondence algorithms.
*/
class StereoMatcher : public Algorithm
{
public:
enum { DISP_SHIFT = 4,
DISP_SCALE = (1 << DISP_SHIFT)
};
/** @brief Computes disparity map for the specified stereo pair
@param left Left 8-bit single-channel image.
@param right Right image of the same size and the same type as the left one.
@param disparity Output disparity map. It has the same size as the input images. Some algorithms,
like StereoBM or StereoSGBM compute 16-bit fixed-point disparity map (where each disparity value
has 4 fractional bits), whereas other algorithms output 32-bit floating-point disparity map.
*/
virtual void compute( InputArray left, InputArray right,
OutputArray disparity ) = 0;
virtual int getMinDisparity() const = 0;
virtual void setMinDisparity(int minDisparity) = 0;
virtual int getNumDisparities() const = 0;
virtual void setNumDisparities(int numDisparities) = 0;
virtual int getBlockSize() const = 0;
virtual void setBlockSize(int blockSize) = 0;
virtual int getSpeckleWindowSize() const = 0;
virtual void setSpeckleWindowSize(int speckleWindowSize) = 0;
virtual int getSpeckleRange() const = 0;
virtual void setSpeckleRange(int speckleRange) = 0;
virtual int getDisp12MaxDiff() const = 0;
virtual void setDisp12MaxDiff(int disp12MaxDiff) = 0;
};
/** @brief Class for computing stereo correspondence using the block matching algorithm, introduced and
contributed to OpenCV by K. Konolige.
*/
class StereoBinaryBM : public StereoMatcher
{
public:
enum { PREFILTER_NORMALIZED_RESPONSE = 0,
PREFILTER_XSOBEL = 1
};
virtual int getPreFilterType() const = 0;
virtual void setPreFilterType(int preFilterType) = 0;
virtual int getPreFilterSize() const = 0;
virtual void setPreFilterSize(int preFilterSize) = 0;
virtual int getPreFilterCap() const = 0;
virtual void setPreFilterCap(int preFilterCap) = 0;
virtual int getTextureThreshold() const = 0;
virtual void setTextureThreshold(int textureThreshold) = 0;
virtual int getUniquenessRatio() const = 0;
virtual void setUniquenessRatio(int uniquenessRatio) = 0;
virtual int getSmallerBlockSize() const = 0;
virtual void setSmallerBlockSize(int blockSize) = 0;
virtual Rect getROI1() const = 0;
virtual void setROI1(Rect roi1) = 0;
virtual Rect getROI2() const = 0;
virtual void setROI2(Rect roi2) = 0;
/** @brief Creates StereoBM object
@param numDisparities the disparity search range. For each pixel algorithm will find the best
disparity from 0 (default minimum disparity) to numDisparities. The search range can then be
shifted by changing the minimum disparity.
@param blockSize the linear size of the blocks compared by the algorithm. The size should be odd
(as the block is centered at the current pixel). Larger block size implies smoother, though less
accurate disparity map. Smaller block size gives more detailed disparity map, but there is higher
chance for algorithm to find a wrong correspondence.
The function create StereoBM object. You can then call StereoBM::compute() to compute disparity for
a specific stereo pair.
*/
CV_EXPORTS static Ptr< cv::stereo::StereoBinaryBM > create(int numDisparities = 0, int blockSize = 21);
};
/** @brief The class implements the modified H. Hirschmuller algorithm @cite HH08 that differs from the original
one as follows:
- By default, the algorithm is single-pass, which means that you consider only 5 directions
instead of 8. Set mode=StereoSGBM::MODE_HH in createStereoSGBM to run the full variant of the
algorithm but beware that it may consume a lot of memory.
- The algorithm matches blocks, not individual pixels. Though, setting blockSize=1 reduces the
blocks to single pixels.
- Mutual information cost function is not implemented. Instead, a simpler Birchfield-Tomasi
sub-pixel metric from @cite BT98 is used. Though, the color images are supported as well.
- Some pre- and post- processing steps from K. Konolige algorithm StereoBM are included, for
example: pre-filtering (StereoBM::PREFILTER_XSOBEL type) and post-filtering (uniqueness
check, quadratic interpolation and speckle filtering).
@note
- (Python) An example illustrating the use of the StereoSGBM matching algorithm can be found
at opencv_source_code/samples/python2/stereo_match.py
*/
class StereoBinarySGBM : public StereoMatcher
{
public:
enum
{
MODE_SGBM = 0,
MODE_HH = 1
};
virtual int getPreFilterCap() const = 0;
virtual void setPreFilterCap(int preFilterCap) = 0;
virtual int getUniquenessRatio() const = 0;
virtual void setUniquenessRatio(int uniquenessRatio) = 0;
virtual int getP1() const = 0;
virtual void setP1(int P1) = 0;
virtual int getP2() const = 0;
virtual void setP2(int P2) = 0;
virtual int getMode() const = 0;
virtual void setMode(int mode) = 0;
/** @brief Creates StereoSGBM object
@param minDisparity Minimum possible disparity value. Normally, it is zero but sometimes
rectification algorithms can shift images, so this parameter needs to be adjusted accordingly.
@param numDisparities Maximum disparity minus minimum disparity. The value is always greater than
zero. In the current implementation, this parameter must be divisible by 16.
@param blockSize Matched block size. It must be an odd number \>=1 . Normally, it should be
somewhere in the 3..11 range.
@param P1 The first parameter controlling the disparity smoothness. See below.
@param P2 The second parameter controlling the disparity smoothness. The larger the values are,
the smoother the disparity is. P1 is the penalty on the disparity change by plus or minus 1
between neighbor pixels. P2 is the penalty on the disparity change by more than 1 between neighbor
pixels. The algorithm requires P2 \> P1 . See stereo_match.cpp sample where some reasonably good
P1 and P2 values are shown (like 8\*number_of_image_channels\*SADWindowSize\*SADWindowSize and
32\*number_of_image_channels\*SADWindowSize\*SADWindowSize , respectively).
@param disp12MaxDiff Maximum allowed difference (in integer pixel units) in the left-right
disparity check. Set it to a non-positive value to disable the check.
@param preFilterCap Truncation value for the prefiltered image pixels. The algorithm first
computes x-derivative at each pixel and clips its value by [-preFilterCap, preFilterCap] interval.
The result values are passed to the Birchfield-Tomasi pixel cost function.
@param uniquenessRatio Margin in percentage by which the best (minimum) computed cost function
value should "win" the second best value to consider the found match correct. Normally, a value
within the 5-15 range is good enough.
@param speckleWindowSize Maximum size of smooth disparity regions to consider their noise speckles
and invalidate. Set it to 0 to disable speckle filtering. Otherwise, set it somewhere in the
50-200 range.
@param speckleRange Maximum disparity variation within each connected component. If you do speckle
filtering, set the parameter to a positive value, it will be implicitly multiplied by 16.
Normally, 1 or 2 is good enough.
@param mode Set it to StereoSGBM::MODE_HH to run the full-scale two-pass dynamic programming
algorithm. It will consume O(W\*H\*numDisparities) bytes, which is large for 640x480 stereo and
huge for HD-size pictures. By default, it is set to false .
The first constructor initializes StereoSGBM with all the default parameters. So, you only have to
set StereoSGBM::numDisparities at minimum. The second constructor enables you to set each parameter
to a custom value.
*/
CV_EXPORTS static Ptr<cv::stereo::StereoBinarySGBM> create(int minDisparity, int numDisparities, int blockSize,
int P1 = 0, int P2 = 0, int disp12MaxDiff = 0,
int preFilterCap = 0, int uniquenessRatio = 0,
int speckleWindowSize = 0, int speckleRange = 0,
int mode = StereoBinarySGBM::MODE_SGBM);
};
//! @}
}//stereo
namespace stereo
{
//! @addtogroup stereo
//! @{
// void correctMatches( InputArray F, InputArray points1, InputArray points2,
// OutputArray newPoints1, OutputArray newPoints2 );
enum {
CV_SPECKLE_REMOVAL_ALGORITHM, CV_SPECKLE_REMOVAL_AVG_ALGORITHM
};
/** @brief Filters off small noise blobs (speckles) in the disparity map
@param img The input 16-bit signed disparity image
@param newVal The disparity value used to paint-off the speckles
@param maxSpeckleSize The maximum speckle size to consider it a speckle. Larger blobs are not
affected by the algorithm
@param maxDiff Maximum difference between neighbor disparity pixels to put them into the same
blob. Note that since StereoBM, StereoSGBM and may be other algorithms return a fixed-point
disparity map, where disparity values are multiplied by 16, this scale factor should be taken into
account when specifying this parameter value.
@param buf The optional temporary buffer to avoid memory allocation within the function.
*/
/** @brief The base class for stereo correspondence algorithms.
*/
class StereoMatcher : public Algorithm
{
public:
enum { DISP_SHIFT = 4,
DISP_SCALE = (1 << DISP_SHIFT)
};
/** @brief Computes disparity map for the specified stereo pair
@param left Left 8-bit single-channel image.
@param right Right image of the same size and the same type as the left one.
@param disparity Output disparity map. It has the same size as the input images. Some algorithms,
like StereoBM or StereoSGBM compute 16-bit fixed-point disparity map (where each disparity value
has 4 fractional bits), whereas other algorithms output 32-bit floating-point disparity map.
*/
virtual void compute( InputArray left, InputArray right,
OutputArray disparity ) = 0;
virtual int getMinDisparity() const = 0;
virtual void setMinDisparity(int minDisparity) = 0;
virtual int getNumDisparities() const = 0;
virtual void setNumDisparities(int numDisparities) = 0;
virtual int getBlockSize() const = 0;
virtual void setBlockSize(int blockSize) = 0;
virtual int getSpeckleWindowSize() const = 0;
virtual void setSpeckleWindowSize(int speckleWindowSize) = 0;
virtual int getSpeckleRange() const = 0;
virtual void setSpeckleRange(int speckleRange) = 0;
virtual int getDisp12MaxDiff() const = 0;
virtual void setDisp12MaxDiff(int disp12MaxDiff) = 0;
};
/** @brief Class for computing stereo correspondence using the block matching algorithm, introduced and
contributed to OpenCV by K. Konolige.
*/
class StereoBinaryBM : public StereoMatcher
{
public:
enum { PREFILTER_NORMALIZED_RESPONSE = 0,
PREFILTER_XSOBEL = 1
};
virtual int getPreFilterType() const = 0;
virtual void setPreFilterType(int preFilterType) = 0;
virtual int getPreFilterSize() const = 0;
virtual void setPreFilterSize(int preFilterSize) = 0;
virtual int getPreFilterCap() const = 0;
virtual void setPreFilterCap(int preFilterCap) = 0;
virtual int getTextureThreshold() const = 0;
virtual void setTextureThreshold(int textureThreshold) = 0;
virtual int getUniquenessRatio() const = 0;
virtual void setUniquenessRatio(int uniquenessRatio) = 0;
virtual int getSmallerBlockSize() const = 0;
virtual void setSmallerBlockSize(int blockSize) = 0;
virtual Rect getROI1() const = 0;
virtual void setROI1(Rect roi1) = 0;
virtual Rect getROI2() const = 0;
virtual void setROI2(Rect roi2) = 0;
virtual int getScalleFactor() const = 0 ;
virtual void setScalleFactor(int factor) = 0;
virtual int getSpekleRemovalTechnique() const = 0 ;
virtual void setSpekleRemovalTechnique(int factor) = 0;
virtual bool getUsePrefilter() const = 0 ;
virtual void setUsePrefilter(bool factor) = 0;
virtual int getBinaryKernelType() const = 0;
virtual void setBinaryKernelType(int value) = 0;
virtual int getAgregationWindowSize() const = 0;
virtual void setAgregationWindowSize(int value) = 0;
/** @brief Creates StereoBM object
@param numDisparities the disparity search range. For each pixel algorithm will find the best
disparity from 0 (default minimum disparity) to numDisparities. The search range can then be
shifted by changing the minimum disparity.
@param blockSize the linear size of the blocks compared by the algorithm. The size should be odd
(as the block is centered at the current pixel). Larger block size implies smoother, though less
accurate disparity map. Smaller block size gives more detailed disparity map, but there is higher
chance for algorithm to find a wrong correspondence.
The function create StereoBM object. You can then call StereoBM::compute() to compute disparity for
a specific stereo pair.
*/
CV_EXPORTS static Ptr< cv::stereo::StereoBinaryBM > create(int numDisparities = 0, int blockSize = 21);
};
/** @brief The class implements the modified H. Hirschmuller algorithm @cite HH08 that differs from the original
one as follows:
- By default, the algorithm is single-pass, which means that you consider only 5 directions
instead of 8. Set mode=StereoSGBM::MODE_HH in createStereoSGBM to run the full variant of the
algorithm but beware that it may consume a lot of memory.
- The algorithm matches blocks, not individual pixels. Though, setting blockSize=1 reduces the
blocks to single pixels.
- Mutual information cost function is not implemented. Instead, a simpler Birchfield-Tomasi
sub-pixel metric from @cite BT98 is used. Though, the color images are supported as well.
- Some pre- and post- processing steps from K. Konolige algorithm StereoBM are included, for
example: pre-filtering (StereoBM::PREFILTER_XSOBEL type) and post-filtering (uniqueness
check, quadratic interpolation and speckle filtering).
@note
- (Python) An example illustrating the use of the StereoSGBM matching algorithm can be found
at opencv_source_code/samples/python2/stereo_match.py
*/
class StereoBinarySGBM : public StereoMatcher
{
public:
enum
{
MODE_SGBM = 0,
MODE_HH = 1
};
virtual int getPreFilterCap() const = 0;
virtual void setPreFilterCap(int preFilterCap) = 0;
virtual int getUniquenessRatio() const = 0;
virtual void setUniquenessRatio(int uniquenessRatio) = 0;
virtual int getP1() const = 0;
virtual void setP1(int P1) = 0;
virtual int getP2() const = 0;
virtual void setP2(int P2) = 0;
virtual int getMode() const = 0;
virtual void setMode(int mode) = 0;
/** @brief Creates StereoSGBM object
@param minDisparity Minimum possible disparity value. Normally, it is zero but sometimes
rectification algorithms can shift images, so this parameter needs to be adjusted accordingly.
@param numDisparities Maximum disparity minus minimum disparity. The value is always greater than
zero. In the current implementation, this parameter must be divisible by 16.
@param blockSize Matched block size. It must be an odd number \>=1 . Normally, it should be
somewhere in the 3..11 range.
@param P1 The first parameter controlling the disparity smoothness. See below.
@param P2 The second parameter controlling the disparity smoothness. The larger the values are,
the smoother the disparity is. P1 is the penalty on the disparity change by plus or minus 1
between neighbor pixels. P2 is the penalty on the disparity change by more than 1 between neighbor
pixels. The algorithm requires P2 \> P1 . See stereo_match.cpp sample where some reasonably good
P1 and P2 values are shown (like 8\*number_of_image_channels\*SADWindowSize\*SADWindowSize and
32\*number_of_image_channels\*SADWindowSize\*SADWindowSize , respectively).
@param disp12MaxDiff Maximum allowed difference (in integer pixel units) in the left-right
disparity check. Set it to a non-positive value to disable the check.
@param preFilterCap Truncation value for the prefiltered image pixels. The algorithm first
computes x-derivative at each pixel and clips its value by [-preFilterCap, preFilterCap] interval.
The result values are passed to the Birchfield-Tomasi pixel cost function.
@param uniquenessRatio Margin in percentage by which the best (minimum) computed cost function
value should "win" the second best value to consider the found match correct. Normally, a value
within the 5-15 range is good enough.
@param speckleWindowSize Maximum size of smooth disparity regions to consider their noise speckles
and invalidate. Set it to 0 to disable speckle filtering. Otherwise, set it somewhere in the
50-200 range.
@param speckleRange Maximum disparity variation within each connected component. If you do speckle
filtering, set the parameter to a positive value, it will be implicitly multiplied by 16.
Normally, 1 or 2 is good enough.
@param mode Set it to StereoSGBM::MODE_HH to run the full-scale two-pass dynamic programming
algorithm. It will consume O(W\*H\*numDisparities) bytes, which is large for 640x480 stereo and
huge for HD-size pictures. By default, it is set to false .
The first constructor initializes StereoSGBM with all the default parameters. So, you only have to
set StereoSGBM::numDisparities at minimum. The second constructor enables you to set each parameter
to a custom value.
*/
CV_EXPORTS static Ptr<cv::stereo::StereoBinarySGBM> create(int minDisparity, int numDisparities, int blockSize,
int P1 = 0, int P2 = 0, int disp12MaxDiff = 0,
int preFilterCap = 0, int uniquenessRatio = 0,
int speckleWindowSize = 0, int speckleRange = 0,
int mode = StereoBinarySGBM::MODE_SGBM);
};
//! @}
}//stereo
} // cv
#ifndef DISABLE_OPENCV_24_COMPATIBILITY
......
modules/stereo/samples/Sample1.cpp 0 → 100644
View file @ df859520
#include <iostream>
#include "opencv2/stereo.hpp"
#include "opencv2/imgproc.hpp"
#include "opencv2/highgui.hpp"
#include <iostream>
using namespace cv;
using namespace stereo;
using namespace std;
//in this example we will load a sequence of images from a file process them and display the result on the screen
//the descriptor used is the modified_census transform
int main(int, char**)
{
//begin the program
cout << " Running Main function \n";
//declare 2 images
Mat image1, image2;
// -- 1. Call the constructor for StereoBinaryBM
int ndisparities = 32; /**< Range of disparity */
int kernelSize = 9; /**< Size of the block window. Must be odd */
Ptr<StereoBinaryBM> sbm = StereoBinaryBM::create(ndisparities, kernelSize);
// -- 2. Set parameters
sbm->setPreFilterCap(31);
sbm->setMinDisparity(0);
sbm->setTextureThreshold(10);
sbm->setUniquenessRatio(0);
sbm->setSpeckleWindowSize(400);//speckle size
sbm->setSpeckleRange(200);
sbm->setDisp12MaxDiff(0);
sbm->setScalleFactor(4);//the scalling factor
sbm->setBinaryKernelType(CV_MODIFIED_CENSUS_TRANSFORM);//binary descriptor kernel
sbm->setAgregationWindowSize(9);
sbm->setSpekleRemovalTechnique(CV_SPECKLE_REMOVAL_AVG_ALGORITHM);//speckle removal algorithm
sbm->setUsePrefilter(false);//prefilter or not the images prior to making the transformations
for(int i = 0 ; i < 200; i++)
{
string path = "D:\\WorkingSec";
string left = "l.bmp";
string right = ".bmp";
std::string s;
std::stringstream out;
out << i;
s = out.str();
string finLeft = path + "\\rezult" + s + left;
string finRigth = path + "\\rezult" + s + right;
image1 = imread(finLeft, CV_8UC1);
image2 = imread(finRigth, CV_8UC1);
//set a certain region of interest
Rect region_of_interest = Rect(0, 20, image1.cols, (image1.rows - 20 - 110));
Mat imgLeft = image1(region_of_interest);
Mat imgRight = image2(region_of_interest);
Mat imgDisparity8U = Mat(imgLeft.rows, imgLeft.cols, CV_8UC1);
if (imgLeft.empty() || imgRight.empty())
{
std::cout << " --(!) Error reading images \n" ; return -1;
}
////-- 3. Calculate the disparity image
sbm->compute(imgLeft, imgRight, imgDisparity8U);
imshow("RealImage", image1);
imshow("Disparity", imgDisparity8U);
waitKey(1);
}
waitKey(0);
return 0;
}
modules/stereo/samples/Sample2.cpp 0 → 100644
View file @ df859520
#include <iostream>
#include "opencv2/stereo.hpp"
#include "opencv2/imgproc.hpp"
#include "opencv2/highgui.hpp"
#include <iostream>
using namespace cv;
using namespace stereo;
using namespace std;
int main(int, char**)
{
//begin the program
cout << " Running Main function \n";
//declare 2 images
Mat image1, image2;
// -- 1. Call the constructor for StereoBinaryBM
int ndisparities = 32; /**< Range of disparity */
int kernelSize = 9; /**< Size of the block window. Must be odd */
Ptr<StereoBinaryBM> sbm = StereoBinaryBM::create(ndisparities, kernelSize);
// -- 2. Set parameters
sbm->setPreFilterCap(31);
sbm->setMinDisparity(0);
sbm->setTextureThreshold(10);
sbm->setUniquenessRatio(0);
sbm->setSpeckleWindowSize(400);//speckle size
sbm->setSpeckleRange(200);
sbm->setDisp12MaxDiff(0);
sbm->setScalleFactor(4);//the scalling factor
sbm->setBinaryKernelType(CV_MEAN_VARIATION);//binary descriptor kernel
sbm->setAgregationWindowSize(9);
sbm->setSpekleRemovalTechnique(CV_SPECKLE_REMOVAL_AVG_ALGORITHM);//speckle removal algorithm
sbm->setUsePrefilter(false);//prefilter or not the images prior to making the transformations
//load 2 images from disc
image1 = imread("D:\\rezult0l.bmp", CV_8UC1);
image2 = imread("D:\\rezult0.bmp", CV_8UC1);
//set a certain region of interest
Rect region_of_interest = Rect(0, 20, image1.cols, (image1.rows - 20 - 110));
Mat imgLeft = image1(region_of_interest);
Mat imgRight = image2(region_of_interest);
Mat imgDisparity8U = Mat(imgLeft.rows, imgLeft.cols, CV_8UC1);
if (imgLeft.empty() || imgRight.empty())
{
std::cout << " --(!) Error reading images \n" ; return -1;
}
////-- 3. Calculate the disparity image
sbm->compute(imgLeft, imgRight, imgDisparity8U);
imshow("RealImage", image1);
imshow("Disparity", imgDisparity8U);
waitKey(0);
return 0;
}
modules/stereo/src/descriptor.hpp
View file @ df859520
......@@ -58,7 +58,7 @@ namespace cv
enum {
CV_DENSE_CENSUS, CV_SPARSE_CENSUS,
CV_CS_CENSUS, CV_MODIFIED_CS_CENSUS, CV_MODIFIED_CENSUS_TRANSFORM,
CV_MEAN_VARIATION
CV_MEAN_VARIATION, CV_STAR_KERNEL
};
//!Mean Variation is a robust kernel that compares a pixel
//!not just with the center but also with the mean of the window
......
modules/stereo/src/matching.cpp
View file @ df859520
......@@ -57,14 +57,12 @@ namespace cv
//maxDisp - represents the maximum disparity
Matching::Matching(int maxDisp, int scalling, int confidence)
{
CV_Assert(maxDisp > 10);
CV_Assert(scalling != 0);
CV_Assert(confidence >= 1);
this->scallingFactor = scalling;
//set the maximum disparity
this->maxDisparity = maxDisp;
setMaxDisparity(maxDisp);
//set scalling factor
setScallingFactor(scalling);
//set the value for the confidence
this->confidenceCheck = confidence;
setConfidence(confidence);
//generate the hamming lut in case SSE is not available
hammingLut();
}
......@@ -82,7 +80,7 @@ namespace cv
void Matching::setScallingFactor(int val)
{
CV_Assert(val > 0);
scallingFactor = val;
this->scallingFactor = val;
}
//!method for getting the scalling factor
int Matching::getScallingFactor()
......@@ -99,10 +97,10 @@ namespace cv
CV_Assert(kernelSize % 2 != 0);
CV_Assert(cost.rows == leftImage.rows);
CV_Assert(cost.cols / (maxDisparity + 1) == leftImage.cols);
// cost.setTo(0);
int *c = (int *)cost.data;
memset(c, 0, sizeof(c[0]) * leftImage.cols * leftImage.rows * (maxDisparity + 1));
parallel_for_(cv::Range(kernelSize / 2,leftImage.rows - kernelSize / 2), hammingDistance(leftImage,rightImage,c,maxDisparity,kernelSize / 2,hamLut));
cost.setTo(0);
//int *c = (int *)cost.data;
//memset(c, 0, sizeof(c[0]) * leftImage.cols * leftImage.rows * (maxDisparity + 1));
parallel_for_(cv::Range(kernelSize / 2,leftImage.rows - kernelSize / 2), hammingDistance(leftImage,rightImage,(int *)cost.data,maxDisparity,kernelSize / 2,hamLut));
}
//preprocessing the cost volume in order to get it ready for aggregation
void Matching::costGathering(const Mat &hammingDistanceCost, Mat &cost)
......@@ -110,12 +108,12 @@ namespace cv
CV_Assert(hammingDistanceCost.rows == hammingDistanceCost.rows);
CV_Assert(hammingDistanceCost.type() == CV_32SC4);
CV_Assert(cost.type() == CV_32SC4);
//cost.setTo(0);
cost.setTo(0);
int maxDisp = maxDisparity;
int width = cost.cols / ( maxDisp + 1) - 1;
int height = cost.rows - 1;
int *c = (int *)cost.data;
memset(c, 0, sizeof(c[0]) * (width + 1) * (height + 1) * (maxDisp + 1));
//memset(c, 0, sizeof(c[0]) * (width + 1) * (height + 1) * (maxDisp + 1));
parallel_for_(cv::Range(1,height), costGatheringHorizontal(hammingDistanceCost,maxDisparity,cost));
for (int i = 1; i <= height; i++)
{
......@@ -136,12 +134,13 @@ namespace cv
CV_Assert(windowSize % 2 != 0);
CV_Assert(partialSums.rows == cost.rows);
CV_Assert(partialSums.cols == cost.cols);
cost.setTo(0);
int win = windowSize / 2;
int *c = (int *)cost.data;
//int *c = (int *)cost.data;
int maxDisp = maxDisparity;
int width = cost.cols / ( maxDisp + 1) - 1;
//int width = cost.cols / ( maxDisp + 1) - 1;
int height = cost.rows - 1;
memset(c, 0, sizeof(c[0]) * width * height * (maxDisp + 1));
//memset(c, 0, sizeof(c[0]) * width * height * (maxDisp + 1));
parallel_for_(cv::Range(win + 1,height - win - 1), agregateCost(partialSums,windowSize,maxDisp,cost));
}
//!Finding the correct disparity from the cost volume, we also make a confidence check
......@@ -303,7 +302,7 @@ namespace cv
void Matching ::setConfidence(double val)
{
CV_Assert(val >= 1);
confidenceCheck = val;
this->confidenceCheck = val;
}
//getter for confidence check
double Matching ::getConfidence()
......@@ -313,26 +312,27 @@ namespace cv
//!Method responsible for generating the disparity map
void Matching::dispartyMapFormation(const Mat &costVolume, Mat &mapFinal, int th)
{
uint8_t *map = mapFinal.data;
mapFinal.setTo(0);
//uint8_t *map = mapFinal.data;
int disparity = maxDisparity;
int width = costVolume.cols / ( disparity + 1) - 1;
//int width = costVolume.cols / ( disparity + 1) - 1;
int height = costVolume.rows - 1;
memset(map, 0, sizeof(map[0]) * width * height);
//memset(map, 0, sizeof(map[0]) * width * height);
parallel_for_(Range(0,height - 1), makeMap(costVolume,th,disparity,confidenceCheck,scallingFactor,mapFinal));
}
//!1x9 median filter
void Matching::Median1x9Filter(const Mat &originalMap, Mat &map)
void Matching::Median1x9Filter(const Mat &originalImage, Mat &filteredImage)
{
CV_Assert(originalMap.rows == map.rows);
CV_Assert(originalMap.cols == map.cols);
parallel_for_(Range(1,originalMap.rows - 2), Median1x9(originalMap,map));
CV_Assert(originalImage.rows == filteredImage.rows);
CV_Assert(originalImage.cols == filteredImage.cols);
parallel_for_(Range(1,originalImage.rows - 2), Median1x9(originalImage,filteredImage));
}
//!9x1 median filter
void Matching::Median9x1Filter(const Mat &originalMap, Mat &map)
void Matching::Median9x1Filter(const Mat &originalImage, Mat &filteredImage)
{
CV_Assert(originalMap.cols == map.cols);
CV_Assert(originalMap.cols == map.cols);
parallel_for_(Range(1,originalMap.cols - 2), Median9x1(originalMap,map));
CV_Assert(originalImage.cols == filteredImage.cols);
CV_Assert(originalImage.cols == filteredImage.cols);
parallel_for_(Range(1,originalImage.cols - 2), Median9x1(originalImage,filteredImage));
}
}
}
modules/stereo/src/matching.hpp
View file @ df859520
......@@ -81,12 +81,12 @@ namespace cv
{
private:
int *left, *right, *c;
int v,kernelSize, width, height;
int v,kernelSize, width, height,_stride;
int MASK;
int *hammLut;
public :
hammingDistance(const Mat &leftImage, const Mat &rightImage, int *cost, int maxDisp, int kerSize, int *hammingLUT):
left((int *)leftImage.data), right((int *)rightImage.data), c(cost), v(maxDisp),kernelSize(kerSize),width(leftImage.cols), height(leftImage.rows), MASK(65535), hammLut(hammingLUT){}
left((int *)leftImage.data), right((int *)rightImage.data), c(cost), v(maxDisp),kernelSize(kerSize),width(leftImage.cols), height(leftImage.rows), _stride((int)leftImage.step1()), MASK(65535), hammLut(hammingLUT){}
void operator()(const cv::Range &r) const {
for (int i = r.start; i <= r.end ; i++)
{
......@@ -246,16 +246,17 @@ namespace cv
class Median1x9:public ParallelLoopBody
{
private:
uint8_t *harta;
uint8_t *mapModified;
int height, width;
uint8_t *original;
uint8_t *filtered;
int height, width,_stride;
public:
Median1x9(const Mat &hartaOriginala, Mat &map)
Median1x9(const Mat &originalImage, Mat &filteredImage)
{
harta = hartaOriginala.data;
mapModified = map.data;
height = hartaOriginala.rows;
width = hartaOriginala.cols;
original = originalImage.data;
filtered = filteredImage.data;
height = originalImage.rows;
width = originalImage.cols;
_stride = (int)originalImage.step;
}
void operator()(const cv::Range &r) const{
for (int m = r.start; m <= r.end; m++)
......@@ -265,7 +266,7 @@ namespace cv
int k = 0;
uint8_t window[9];
for (int i = n - 4; i <= n + 4; ++i)
window[k++] = harta[m * width + i];
window[k++] = original[m * _stride + i];
for (int j = 0; j < 5; ++j)
{
int min = j;
......@@ -276,7 +277,7 @@ namespace cv
window[j] = window[min];
window[min] = temp;
}
mapModified[m * width + n] = window[4];
filtered[m * _stride + n] = window[4];
}
}
}
......@@ -285,16 +286,17 @@ namespace cv
class Median9x1:public ParallelLoopBody
{
private:
uint8_t *harta;
uint8_t *mapModified;
int height, width;
uint8_t *original;
uint8_t *filtered;
int height, width, _stride;
public:
Median9x1(const Mat &hartaOriginala, Mat &map)
Median9x1(const Mat &originalImage, Mat &filteredImage)
{
harta = hartaOriginala.data;
mapModified = map.data;
height = hartaOriginala.rows;
width = hartaOriginala.cols;
original = originalImage.data;
filtered = filteredImage.data;
height = originalImage.rows;
width = originalImage.cols;
_stride = (int)originalImage.step;
}
void operator()(const Range &r) const{
for (int n = r.start; n <= r.end; ++n)
......@@ -304,7 +306,7 @@ namespace cv
int k = 0;
uint8_t window[9];
for (int i = m - 4; i <= m + 4; ++i)
window[k++] = harta[i * width + n];
window[k++] = original[i * _stride + n];
for (int j = 0; j < 5; j++)
{
int min = j;
......@@ -315,12 +317,12 @@ namespace cv
window[j] = window[min];
window[min] = temp;
}
mapModified[m * width + n] = window[4];
filtered[m * _stride + n] = window[4];
}
}
}
};
public:
protected:
//!method for setting the maximum disparity
void setMaxDisparity(int val);
//!method for getting the disparity
......@@ -347,12 +349,13 @@ namespace cv
*th - is the LR threshold
*/
void dispartyMapFormation(const Mat &costVolume, Mat &map, int th);
void smallRegionRemoval(const Mat &input, int t, Mat &out);
public:
static void Median1x9Filter(const Mat &inputImage, Mat &outputImage);
static void Median9x1Filter(const Mat &inputImage, Mat &outputImage);
//!constructor for the matching class
//!maxDisp - represents the maximum disparity
//!a median filter that has proven to work a bit better especially when applied on disparity maps
static void Median1x9Filter(const Mat &hartaOriginala, Mat &map);
static void Median9x1Filter(const Mat &hartaOriginala, Mat &map);
void smallRegionRemoval(const Mat &harta, int t, Mat &out);
Matching(int maxDisp, int scallingFactor = 4,int confidenceCheck = 6);
Matching(void);
~Matching(void);
......
modules/stereo/src/stereo_binary_bm.cpp
View file @ df859520
......@@ -46,6 +46,8 @@
\****************************************************************************************/
#include "precomp.hpp"
#include "descriptor.hpp"
#include "matching.hpp"
#include <stdio.h>
#include <limits>
......@@ -53,14 +55,15 @@ namespace cv
{
namespace stereo
{
struct StereoBinaryBMParams
{
StereoBinaryBMParams(int _numDisparities = 64, int _SADWindowSize = 9)
StereoBinaryBMParams(int _numDisparities = 64, int _kernelSize = 9)
{
preFilterType = StereoBinaryBM::PREFILTER_XSOBEL;
preFilterSize = 9;
preFilterCap = 31;
SADWindowSize = _SADWindowSize;
kernelSize = _kernelSize;
minDisparity = 0;
numDisparities = _numDisparities > 0 ? _numDisparities : 64;
textureThreshold = 10;
......@@ -69,12 +72,17 @@ namespace cv
roi1 = roi2 = Rect(0, 0, 0, 0);
disp12MaxDiff = -1;
dispType = CV_16S;
usePrefilter = false;
regionRemoval = 1;
scalling = 4;
kernelType = CV_MODIFIED_CENSUS_TRANSFORM;
agregationWindowSize = 9;
}
int preFilterType;
int preFilterSize;
int preFilterCap;
int SADWindowSize;
int kernelSize;
int minDisparity;
int numDisparities;
int textureThreshold;
......@@ -84,6 +92,11 @@ namespace cv
Rect roi1, roi2;
int disp12MaxDiff;
int dispType;
int scalling;
bool usePrefilter;
int regionRemoval;
int kernelType;
int agregationWindowSize;
};
static void prefilterNorm(const Mat& src, Mat& dst, int winsize, int ftzero, uchar* buf)
......@@ -104,7 +117,6 @@ namespace cv
for (x = 0; x < size.width; x++)
vsum[x] = (ushort)(sptr[x] * (wsz2 + 2));
for (y = 1; y < wsz2; y++)
{
for (x = 0; x < size.width; x++)
......@@ -228,179 +240,6 @@ namespace cv
static const int DISPARITY_SHIFT = 4;
static void
findStereoCorrespondenceBM(const Mat& left, const Mat& right,
Mat& disp, Mat& cost, const StereoBinaryBMParams& state,
uchar* buf, int _dy0, int _dy1)
{
const int ALIGN = 16;
int x, y, d;
int wsz = state.SADWindowSize, wsz2 = wsz / 2;
int dy0 = MIN(_dy0, wsz2 + 1), dy1 = MIN(_dy1, wsz2 + 1);
int ndisp = state.numDisparities;
int mindisp = state.minDisparity;
int lofs = MAX(ndisp - 1 + mindisp, 0);
int rofs = -MIN(ndisp - 1 + mindisp, 0);
int width = left.cols, height = left.rows;
int width1 = width - rofs - ndisp + 1;
int ftzero = state.preFilterCap;
int textureThreshold = state.textureThreshold;
int uniquenessRatio = state.uniquenessRatio;
short FILTERED = (short)((mindisp - 1) << DISPARITY_SHIFT);
int *sad, *hsad0, *hsad, *hsad_sub, *htext;
uchar *cbuf0, *cbuf;
const uchar* lptr0 = left.ptr() + lofs;
const uchar* rptr0 = right.ptr() + rofs;
const uchar *lptr, *lptr_sub, *rptr;
short* dptr = disp.ptr<short>();
int sstep = (int)left.step;
int dstep = (int)(disp.step / sizeof(dptr[0]));
int cstep = (height + dy0 + dy1)*ndisp;
int costbuf = 0;
int coststep = cost.data ? (int)(cost.step / sizeof(costbuf)) : 0;
const int TABSZ = 256;
uchar tab[TABSZ];
sad = (int*)alignPtr(buf + sizeof(sad[0]), ALIGN);
hsad0 = (int*)alignPtr(sad + ndisp + 1 + dy0*ndisp, ALIGN);
htext = (int*)alignPtr((int*)(hsad0 + (height + dy1)*ndisp) + wsz2 + 2, ALIGN);
cbuf0 = (uchar*)alignPtr((uchar*)(htext + height + wsz2 + 2) + dy0*ndisp, ALIGN);
for (x = 0; x < TABSZ; x++)
tab[x] = (uchar)std::abs(x - ftzero);
// initialize buffers
memset(hsad0 - dy0*ndisp, 0, (height + dy0 + dy1)*ndisp*sizeof(hsad0[0]));
memset(htext - wsz2 - 1, 0, (height + wsz + 1)*sizeof(htext[0]));
for (x = -wsz2 - 1; x < wsz2; x++)
{
hsad = hsad0 - dy0*ndisp; cbuf = cbuf0 + (x + wsz2 + 1)*cstep - dy0*ndisp;
lptr = lptr0 + std::min(std::max(x, -lofs), width - lofs - 1) - dy0*sstep;
rptr = rptr0 + std::min(std::max(x, -rofs), width - rofs - 1) - dy0*sstep;
for (y = -dy0; y < height + dy1; y++, hsad += ndisp, cbuf += ndisp, lptr += sstep, rptr += sstep)
{
int lval = lptr[0];
for (d = 0; d < ndisp; d++)
{
int diff = std::abs(lval - rptr[d]);
cbuf[d] = (uchar)diff;
hsad[d] = (int)(hsad[d] + diff);
}
htext[y] += tab[lval];
}
}
// initialize the left and right borders of the disparity map
for (y = 0; y < height; y++)
{
for (x = 0; x < lofs; x++)
dptr[y*dstep + x] = FILTERED;
for (x = lofs + width1; x < width; x++)
dptr[y*dstep + x] = FILTERED;
}
dptr += lofs;
for (x = 0; x < width1; x++, dptr++)
{
int* costptr = cost.data ? cost.ptr<int>() + lofs + x : &costbuf;
int x0 = x - wsz2 - 1, x1 = x + wsz2;
const uchar* cbuf_sub = cbuf0 + ((x0 + wsz2 + 1) % (wsz + 1))*cstep - dy0*ndisp;
cbuf = cbuf0 + ((x1 + wsz2 + 1) % (wsz + 1))*cstep - dy0*ndisp;
hsad = hsad0 - dy0*ndisp;
lptr_sub = lptr0 + MIN(MAX(x0, -lofs), width - 1 - lofs) - dy0*sstep;
lptr = lptr0 + MIN(MAX(x1, -lofs), width - 1 - lofs) - dy0*sstep;
rptr = rptr0 + MIN(MAX(x1, -rofs), width - 1 - rofs) - dy0*sstep;
for (y = -dy0; y < height + dy1; y++, cbuf += ndisp, cbuf_sub += ndisp,
hsad += ndisp, lptr += sstep, lptr_sub += sstep, rptr += sstep)
{
int lval = lptr[0];
for (d = 0; d < ndisp; d++)
{
int diff = std::abs(lval - rptr[d]);
cbuf[d] = (uchar)diff;
hsad[d] = hsad[d] + diff - cbuf_sub[d];
}
htext[y] += tab[lval] - tab[lptr_sub[0]];
}
// fill borders
for (y = dy1; y <= wsz2; y++)
htext[height + y] = htext[height + dy1 - 1];
for (y = -wsz2 - 1; y < -dy0; y++)
htext[y] = htext[-dy0];
// initialize sums
int tsum = 0;
{
for (d = 0; d < ndisp; d++)
sad[d] = (int)(hsad0[d - ndisp*dy0] * (wsz2 + 2 - dy0));
hsad = hsad0 + (1 - dy0)*ndisp;
for (y = 1 - dy0; y < wsz2; y++, hsad += ndisp)
for (d = 0; d < ndisp; d++)
sad[d] = (int)(sad[d] + hsad[d]);
for (y = -wsz2 - 1; y < wsz2; y++)
tsum += htext[y];
}
// finally, start the real processing
{
for (y = 0; y < height; y++)
{
int minsad = INT_MAX, mind = -1;
hsad = hsad0 + MIN(y + wsz2, height + dy1 - 1)*ndisp;
hsad_sub = hsad0 + MAX(y - wsz2 - 1, -dy0)*ndisp;
for (d = 0; d < ndisp; d++)
{
int currsad = sad[d] + hsad[d] - hsad_sub[d];
sad[d] = currsad;
if (currsad < minsad)
{
minsad = currsad;
mind = d;
}
}
tsum += htext[y + wsz2] - htext[y - wsz2 - 1];
if (tsum < textureThreshold)
{
dptr[y*dstep] = FILTERED;
continue;
}
if (uniquenessRatio > 0)
{
int thresh = minsad + (minsad * uniquenessRatio / 100);
for (d = 0; d < ndisp; d++)
{
if ((d < mind - 1 || d > mind + 1) && sad[d] <= thresh)
break;
}
if (d < ndisp)
{
dptr[y*dstep] = FILTERED;
continue;
}
}
{
sad[-1] = sad[1];
sad[ndisp] = sad[ndisp - 2];
int p = sad[mind + 1], n = sad[mind - 1];
d = p + n - 2 * sad[mind] + std::abs(p - n);
dptr[y*dstep] = (short)(((ndisp - mind - 1 + mindisp) * 256 + (d != 0 ? (p - n) * 256 / d : 0) + 15) >> 4);
costptr[y*coststep] = sad[mind];
}
}
}
}
}
struct PrefilterInvoker : public ParallelLoopBody
{
PrefilterInvoker(const Mat& left0, const Mat& right0, Mat& left, Mat& right,
......@@ -429,83 +268,7 @@ namespace cv
StereoBinaryBMParams* state;
};
struct FindStereoCorrespInvoker : public ParallelLoopBody
{
FindStereoCorrespInvoker(const Mat& _left, const Mat& _right,
Mat& _disp, StereoBinaryBMParams* _state,
int _nstripes, size_t _stripeBufSize,
bool _useShorts, Rect _validDisparityRect,
Mat& _slidingSumBuf, Mat& _cost)
{
left = &_left; right = &_right;
disp = &_disp; state = _state;
nstripes = _nstripes; stripeBufSize = _stripeBufSize;
useShorts = _useShorts;
validDisparityRect = _validDisparityRect;
slidingSumBuf = &_slidingSumBuf;
cost = &_cost;
}
void operator()(const Range& range) const
{
int cols = left->cols, rows = left->rows;
int _row0 = std::min(cvRound(range.start * rows / nstripes), rows);
int _row1 = std::min(cvRound(range.end * rows / nstripes), rows);
uchar *ptr = slidingSumBuf->ptr() + range.start * stripeBufSize;
int FILTERED = (state->minDisparity - 1) * 16;
Rect roi = validDisparityRect & Rect(0, _row0, cols, _row1 - _row0);
if (roi.height == 0)
return;
int row0 = roi.y;
int row1 = roi.y + roi.height;
Mat part;
if (row0 > _row0)
{
part = disp->rowRange(_row0, row0);
part = Scalar::all(FILTERED);
}
if (_row1 > row1)
{
part = disp->rowRange(row1, _row1);
part = Scalar::all(FILTERED);
}
Mat left_i = left->rowRange(row0, row1);
Mat right_i = right->rowRange(row0, row1);
Mat disp_i = disp->rowRange(row0, row1);
Mat cost_i = state->disp12MaxDiff >= 0 ? cost->rowRange(row0, row1) : Mat();
findStereoCorrespondenceBM(left_i, right_i, disp_i, cost_i, *state, ptr, row0, rows - row1);
if (state->disp12MaxDiff >= 0)
validateDisparity(disp_i, cost_i, state->minDisparity, state->numDisparities, state->disp12MaxDiff);
if (roi.x > 0)
{
part = disp_i.colRange(0, roi.x);
part = Scalar::all(FILTERED);
}
if (roi.x + roi.width < cols)
{
part = disp_i.colRange(roi.x + roi.width, cols);
part = Scalar::all(FILTERED);
}
}
protected:
const Mat *left, *right;
Mat* disp, *slidingSumBuf, *cost;
StereoBinaryBMParams *state;
int nstripes;
size_t stripeBufSize;
bool useShorts;
Rect validDisparityRect;
};
class StereoBinaryBMImpl : public StereoBinaryBM
class StereoBinaryBMImpl : public StereoBinaryBM,public Matching
{
public:
StereoBinaryBMImpl()
......@@ -513,9 +276,9 @@ namespace cv
params = StereoBinaryBMParams();
}
StereoBinaryBMImpl(int _numDisparities, int _SADWindowSize)
StereoBinaryBMImpl(int _numDisparities, int _kernelSize) : Matching(_numDisparities)
{
params = StereoBinaryBMParams(_numDisparities, _SADWindowSize);
params = StereoBinaryBMParams(_numDisparities, _kernelSize);
}
void compute(InputArray leftarr, InputArray rightarr, OutputArray disparr)
......@@ -542,9 +305,9 @@ namespace cv
if (params.preFilterCap < 1 || params.preFilterCap > 63)
CV_Error(Error::StsOutOfRange, "preFilterCap must be within 1..63");
if (params.SADWindowSize < 5 || params.SADWindowSize > 255 || params.SADWindowSize % 2 == 0 ||
params.SADWindowSize >= std::min(leftsize.width, leftsize.height))
CV_Error(Error::StsOutOfRange, "SADWindowSize must be odd, be within 5..255 and be not larger than image width or height");
if (params.kernelSize < 5 || params.kernelSize > 255 || params.kernelSize % 2 == 0 ||
params.kernelSize >= std::min(leftsize.width, leftsize.height))
CV_Error(Error::StsOutOfRange, "kernelSize must be odd, be within 5..255 and be not larger than image width or height");
if (params.numDisparities <= 0 || params.numDisparities % 16 != 0)
CV_Error(Error::StsOutOfRange, "numDisparities must be positive and divisble by 16");
......@@ -557,85 +320,111 @@ namespace cv
int FILTERED = (params.minDisparity - 1) << DISPARITY_SHIFT;
Mat left0 = leftarr.getMat(), right0 = rightarr.getMat();
disparr.create(left0.size(), dtype);
Mat disp0 = disparr.getMat();
censusImage[0].create(left0.rows,left0.cols,CV_32SC4);
censusImage[1].create(left0.rows,left0.cols,CV_32SC4);
partialSumsLR.create(left0.rows + 1,(left0.cols + 1) * (params.numDisparities + 1),CV_32SC4);
agregatedHammingLRCost.create(left0.rows + 1,(left0.cols + 1) * (params.numDisparities + 1),CV_32SC4);
hammingDistance.create(left0.rows, left0.cols * (params.numDisparities + 1),CV_32SC4);
preFilteredImg0.create(left0.size(), CV_8U);
preFilteredImg1.create(left0.size(), CV_8U);
cost.create(left0.size(), CV_16S);
Mat left = preFilteredImg0, right = preFilteredImg1;
int mindisp = params.minDisparity;
int ndisp = params.numDisparities;
int width = left0.cols;
int height = left0.rows;
int lofs = std::max(ndisp - 1 + mindisp, 0);
int rofs = -std::min(ndisp - 1 + mindisp, 0);
int width1 = width - rofs - ndisp + 1;
if (lofs >= width || rofs >= width || width1 < 1)
{
disp0 = Scalar::all(FILTERED * (disp0.type() < CV_32F ? 1 : 1. / (1 << DISPARITY_SHIFT)));
return;
}
Mat disp = disp0;
if (dtype == CV_32F)
{
dispbuf.create(disp0.size(), CV_16S);
disp = dispbuf;
}
int wsz = params.SADWindowSize;
int wsz = params.kernelSize;
int bufSize0 = (int)((ndisp + 2)*sizeof(int));
bufSize0 += (int)((height + wsz + 2)*ndisp*sizeof(int));
bufSize0 += (int)((height + wsz + 2)*sizeof(int));
bufSize0 += (int)((height + wsz + 2)*ndisp*(wsz + 2)*sizeof(uchar) + 256);
int bufSize1 = (int)((width + params.preFilterSize + 2) * sizeof(int) + 256);
int bufSize2 = 0;
if (params.speckleRange >= 0 && params.speckleWindowSize > 0)
bufSize2 = width*height*(sizeof(Point_<short>) + sizeof(int) + sizeof(uchar));
#if CV_SSE2
bool useShorts = params.preFilterCap <= 31 && params.SADWindowSize <= 21 && checkHardwareSupport(CV_CPU_SSE2);
#else
const bool useShorts = false;
#endif
const double SAD_overhead_coeff = 10.0;
double N0 = 8000000 / (useShorts ? 1 : 4); // approx tbb's min number instructions reasonable for one thread
double maxStripeSize = std::min(std::max(N0 / (width * ndisp), (wsz - 1) * SAD_overhead_coeff), (double)height);
int nstripes = cvCeil(height / maxStripeSize);
int bufSize = std::max(bufSize0 * nstripes, std::max(bufSize1 * 2, bufSize2));
if (slidingSumBuf.cols < bufSize)
slidingSumBuf.create(1, bufSize, CV_8U);
uchar *_buf = slidingSumBuf.ptr();
if(params.usePrefilter == true)
{
uchar *_buf = slidingSumBuf.ptr();
parallel_for_(Range(0, 2), PrefilterInvoker(left0, right0, left, right, _buf, _buf + bufSize1, &params), 1);
parallel_for_(Range(0, 2), PrefilterInvoker(left0, right0, left, right, _buf, _buf + bufSize1, &params), 1);
}
else if(params.usePrefilter == false)
{
left = left0;
right = right0;
}
if(params.kernelType == CV_SPARSE_CENSUS)
{
censusTransform(left,right,params.kernelSize,censusImage[0],censusImage[1],CV_SPARSE_CENSUS);
}
else if(params.kernelType == CV_DENSE_CENSUS)
{
censusTransform(left,right,params.kernelSize,censusImage[0],censusImage[1],CV_SPARSE_CENSUS);
}
else if(params.kernelType == CV_CS_CENSUS)
{
symetricCensusTransform(left,right,params.kernelSize,censusImage[0],censusImage[1],CV_CS_CENSUS);
}
else if(params.kernelType == CV_MODIFIED_CS_CENSUS)
{
symetricCensusTransform(left,right,params.kernelSize,censusImage[0],censusImage[1],CV_MODIFIED_CS_CENSUS);
}
else if(params.kernelType == CV_MODIFIED_CENSUS_TRANSFORM)
{
modifiedCensusTransform(left,right,params.kernelSize,censusImage[0],censusImage[1],CV_MODIFIED_CENSUS_TRANSFORM,0);
}
else if(params.kernelType == CV_MEAN_VARIATION)
{
parSumsIntensityImage[0].create(left0.rows, left0.cols,CV_32SC4);
parSumsIntensityImage[1].create(left0.rows, left0.cols,CV_32SC4);
Integral[0].create(left0.rows,left0.cols,CV_32SC4);
Integral[1].create(left0.rows,left0.cols,CV_32SC4);
integral(left, parSumsIntensityImage[0],CV_32S);
integral(right, parSumsIntensityImage[1],CV_32S);
imageMeanKernelSize(parSumsIntensityImage[0], params.kernelSize,Integral[0]);
imageMeanKernelSize(parSumsIntensityImage[1], params.kernelSize, Integral[1]);
modifiedCensusTransform(left,right,params.kernelSize,censusImage[0],censusImage[1],CV_MEAN_VARIATION,0,Integral[0], Integral[1]);
}
else if(params.kernelType == CV_STAR_KERNEL)
{
starCensusTransform(left,right,params.kernelSize,censusImage[0],censusImage[1]);
}
hammingDistanceBlockMatching(censusImage[0], censusImage[1], hammingDistance);
costGathering(hammingDistance, partialSumsLR);
blockAgregation(partialSumsLR, params.agregationWindowSize, agregatedHammingLRCost);
dispartyMapFormation(agregatedHammingLRCost, disp0, 3);
Median1x9Filter(disp0, disp0);
Median9x1Filter(disp0,disp0);
if(params.regionRemoval == CV_SPECKLE_REMOVAL_AVG_ALGORITHM)
{
smallRegionRemoval(disp0,params.speckleWindowSize,disp0);
}
else if(params.regionRemoval == CV_SPECKLE_REMOVAL_ALGORITHM)
{
if (params.speckleRange >= 0 && params.speckleWindowSize > 0)
filterSpeckles(disp0, FILTERED, params.speckleWindowSize, params.speckleRange, slidingSumBuf);
}
}
int getAgregationWindowSize() const { return params.agregationWindowSize;}
void setAgregationWindowSize(int value = 9) { params.agregationWindowSize = value;}
Rect validDisparityRect(0, 0, width, height), R1 = params.roi1, R2 = params.roi2;
validDisparityRect = getValidDisparityROI(R1.area() > 0 ? Rect(0, 0, width, height) : validDisparityRect,
R2.area() > 0 ? Rect(0, 0, width, height) : validDisparityRect,
params.minDisparity, params.numDisparities,
params.SADWindowSize);
int getBinaryKernelType() const { return params.kernelType;}
void setBinaryKernelType(int value = CV_MODIFIED_CENSUS_TRANSFORM) { params.kernelType = value; }
parallel_for_(Range(0, nstripes),
FindStereoCorrespInvoker(left, right, disp, &params, nstripes,
bufSize0, useShorts, validDisparityRect,
slidingSumBuf, cost));
int getSpekleRemovalTechnique() const { return params.regionRemoval;}
void setSpekleRemovalTechnique(int factor = CV_SPECKLE_REMOVAL_AVG_ALGORITHM) { params.regionRemoval = factor; }
if (params.speckleRange >= 0 && params.speckleWindowSize > 0)
filterSpeckles(disp, FILTERED, params.speckleWindowSize, params.speckleRange, slidingSumBuf);
bool getUsePrefilter() const { return params.usePrefilter;}
void setUsePrefilter(bool value = false) { params.usePrefilter = value;}
if (disp0.data != disp.data)
disp.convertTo(disp0, disp0.type(), 1. / (1 << DISPARITY_SHIFT), 0);
}
int getScalleFactor() const { return params.scalling;}
void setScalleFactor(int factor) {params.scalling = factor; setScallingFactor(factor);}
int getMinDisparity() const { return params.minDisparity; }
void setMinDisparity(int minDisparity) { params.minDisparity = minDisparity; }
......@@ -643,8 +432,8 @@ namespace cv
int getNumDisparities() const { return params.numDisparities; }
void setNumDisparities(int numDisparities) { params.numDisparities = numDisparities; }
int getBlockSize() const { return params.SADWindowSize; }
void setBlockSize(int blockSize) { params.SADWindowSize = blockSize; }
int getBlockSize() const { return params.kernelSize; }
void setBlockSize(int blockSize) { params.kernelSize = blockSize; }
int getSpeckleWindowSize() const { return params.speckleWindowSize; }
void setSpeckleWindowSize(int speckleWindowSize) { params.speckleWindowSize = speckleWindowSize; }
......@@ -684,7 +473,7 @@ namespace cv
fs << "name" << name_
<< "minDisparity" << params.minDisparity
<< "numDisparities" << params.numDisparities
<< "blockSize" << params.SADWindowSize
<< "blockSize" << params.kernelSize
<< "speckleWindowSize" << params.speckleWindowSize
<< "speckleRange" << params.speckleRange
<< "disp12MaxDiff" << params.disp12MaxDiff
......@@ -701,7 +490,7 @@ namespace cv
CV_Assert(n.isString() && String(n) == name_);
params.minDisparity = (int)fn["minDisparity"];
params.numDisparities = (int)fn["numDisparities"];
params.SADWindowSize = (int)fn["blockSize"];
params.kernelSize = (int)fn["blockSize"];
params.speckleWindowSize = (int)fn["speckleWindowSize"];
params.speckleRange = (int)fn["speckleRange"];
params.disp12MaxDiff = (int)fn["disp12MaxDiff"];
......@@ -716,15 +505,21 @@ namespace cv
StereoBinaryBMParams params;
Mat preFilteredImg0, preFilteredImg1, cost, dispbuf;
Mat slidingSumBuf;
Mat parSumsIntensityImage[2];
Mat censusImage[2];
Mat hammingDistance;
Mat partialSumsLR;
Mat agregatedHammingLRCost;
Mat Integral[2];
static const char* name_;
};
const char* StereoBinaryBMImpl::name_ = "StereoMatcher.BM";
Ptr<StereoBinaryBM> StereoBinaryBM::create(int _numDisparities, int _SADWindowSize)
Ptr<StereoBinaryBM> StereoBinaryBM::create(int _numDisparities, int _kernelSize)
{
return makePtr<StereoBinaryBMImpl>(_numDisparities, _SADWindowSize);
return makePtr<StereoBinaryBMImpl>(_numDisparities, _kernelSize);
}
}
}
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment