generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Mantiuk's tonemapping

Browse Source
This commit is contained in:
Fedor Morozov 2013-08-05 19:22:42 +04:00
parent c51b50b44c
commit 17609b90c7
9 changed files with 835 additions and 630 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
modules/highgui/src/grfmt_hdr.cpp
View File

@ -123,10 +123,9 @@ HdrEncoder::~HdrEncoder()
bool HdrEncoder::write( const Mat& _img, const std::vector<int>& params )
{
CV_Assert(_img.channels() == 3);
Mat img;
if(_img.depth() == CV_32F) {
_img.convertTo(img, CV_32FC3);
} else {
if(_img.depth() != CV_32F) {
_img.convertTo(img, CV_32FC3, 1/255.0f);
}
CV_Assert(params.empty() || params[0] == HDR_NONE || params[0] == HDR_RLE);

117
modules/photo/include/opencv2/photo.hpp
View File

@ -87,8 +87,8 @@ class CV_EXPORTS_W Tonemap : public Algorithm
public:
CV_WRAP virtual void process(InputArray src, OutputArray dst) = 0;
CV_WRAP virtual float getGamma() const = 0;
CV_WRAP virtual void setGamma(float gamma) = 0;
CV_WRAP virtual float getGamma() const = 0;
CV_WRAP virtual void setGamma(float gamma) = 0;
};
class CV_EXPORTS_W TonemapLinear : public Tonemap
@ -102,71 +102,92 @@ CV_EXPORTS_W Ptr<TonemapLinear> createTonemapLinear(float gamma = 1.0f);
class CV_EXPORTS_W TonemapDrago : public Tonemap
{
public:
CV_WRAP virtual float getBias() const = 0;
CV_WRAP virtual void setBias(float bias) = 0;
CV_WRAP virtual float getSaturation() const = 0;
CV_WRAP virtual void setSaturation(float saturation) = 0;
CV_WRAP virtual float getBias() const = 0;
CV_WRAP virtual void setBias(float bias) = 0;
};
CV_EXPORTS_W Ptr<TonemapDrago> createTonemapDrago(float gamma = 1.0f, float bias = 0.85f);
CV_EXPORTS_W Ptr<TonemapDrago> createTonemapDrago(float gamma = 1.0f, float saturation = 1.0f, float bias = 0.85f);
// "Fast Bilateral Filtering for the Display of High-Dynamic-Range Images", Durand, Dorsey, 2002
class CV_EXPORTS_W TonemapDurand : public Tonemap
{
public:
CV_WRAP virtual float getContrast() const = 0;
CV_WRAP virtual void setContrast(float contrast) = 0;
CV_WRAP virtual float getSigmaSpace() const = 0;
CV_WRAP virtual void setSigmaSpace(float sigma_space) = 0;
CV_WRAP virtual float getSaturation() const = 0;
CV_WRAP virtual void setSaturation(float saturation) = 0;
CV_WRAP virtual float getSigmaColor() const = 0;
CV_WRAP virtual void setSigmaColor(float sigma_color) = 0;
CV_WRAP virtual float getContrast() const = 0;
CV_WRAP virtual void setContrast(float contrast) = 0;
CV_WRAP virtual float getSigmaSpace() const = 0;
CV_WRAP virtual void setSigmaSpace(float sigma_space) = 0;
CV_WRAP virtual float getSigmaColor() const = 0;
CV_WRAP virtual void setSigmaColor(float sigma_color) = 0;
};
CV_EXPORTS_W Ptr<TonemapDurand>
createTonemapDurand(float gamma = 1.0f, float contrast = 4.0f, float sigma_space = 2.0f, float sigma_color = 2.0f);
createTonemapDurand(float gamma = 1.0f, float saturation = 1.0f, float contrast = 4.0f, float sigma_space = 2.0f, float sigma_color = 2.0f);
// "Dynamic Range Reduction Inspired by Photoreceptor Physiology", Reinhard, Devlin, 2005
class CV_EXPORTS_W TonemapReinhardDevlin : public Tonemap
{
public:
CV_WRAP virtual float getIntensity() const = 0;
CV_WRAP virtual void setIntensity(float intensity) = 0;
CV_WRAP virtual float getIntensity() const = 0;
CV_WRAP virtual void setIntensity(float intensity) = 0;
CV_WRAP virtual float getLightAdaptation() const = 0;
CV_WRAP virtual void setLightAdaptation(float light_adapt) = 0;
CV_WRAP virtual float getLightAdaptation() const = 0;
CV_WRAP virtual void setLightAdaptation(float light_adapt) = 0;
CV_WRAP virtual float getColorAdaptation() const = 0;
CV_WRAP virtual void setColorAdaptation(float color_adapt) = 0;
CV_WRAP virtual float getColorAdaptation() const = 0;
CV_WRAP virtual void setColorAdaptation(float color_adapt) = 0;
};
CV_EXPORTS_W Ptr<TonemapReinhardDevlin>
createTonemapReinhardDevlin(float gamma = 1.0f, float intensity = 0.0f, float light_adapt = 1.0f, float color_adapt = 0.0f);
class CV_EXPORTS_W TonemapMantiuk : public Tonemap
{
public:
CV_WRAP virtual float getScale() const = 0;
CV_WRAP virtual void setScale(float scale) = 0;
CV_WRAP virtual float getSaturation() const = 0;
CV_WRAP virtual void setSaturation(float saturation) = 0;
};
CV_EXPORTS_W Ptr<TonemapMantiuk>
createTonemapMantiuk(float gamma = 1.0f, float scale = 0.7f, float saturation = 1.0f);
class CV_EXPORTS_W ExposureAlign : public Algorithm
{
public:
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArrayOfArrays dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArrayOfArrays dst,
const std::vector<float>& times, InputArray response) = 0;
};
class CV_EXPORTS_W AlignMTB : public ExposureAlign
{
public:
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArrayOfArrays dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArrayOfArrays dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArrayOfArrays dst) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArrayOfArrays dst) = 0;
CV_WRAP virtual void calculateShift(InputArray img0, InputArray img1, Point& shift) = 0;
CV_WRAP virtual void shiftMat(InputArray src, OutputArray dst, const Point shift) = 0;
CV_WRAP virtual void calculateShift(InputArray img0, InputArray img1, Point& shift) = 0;
CV_WRAP virtual void shiftMat(InputArray src, OutputArray dst, const Point shift) = 0;
CV_WRAP virtual int getMaxBits() const = 0;
CV_WRAP virtual void setMaxBits(int max_bits) = 0;
CV_WRAP virtual int getMaxBits() const = 0;
CV_WRAP virtual void setMaxBits(int max_bits) = 0;
CV_WRAP virtual int getExcludeRange() const = 0;
CV_WRAP virtual void setExcludeRange(int exclude_range) = 0;
CV_WRAP virtual int getExcludeRange() const = 0;
CV_WRAP virtual void setExcludeRange(int exclude_range) = 0;
};
// "Fast, Robust Image Registration for Compositing High Dynamic Range Photographs from Handheld Exposures", Ward, 2003
@ -176,7 +197,7 @@ CV_EXPORTS_W Ptr<AlignMTB> createAlignMTB(int max_bits = 6, int exclude_range =
class CV_EXPORTS_W ExposureCalibrate : public Algorithm
{
public:
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst, std::vector<float>& times) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst, std::vector<float>& times) = 0;
};
// "Recovering High Dynamic Range Radiance Maps from Photographs", Debevec, Malik, 1997
@ -184,11 +205,11 @@ public:
class CV_EXPORTS_W CalibrateDebevec : public ExposureCalibrate
{
public:
CV_WRAP virtual float getLambda() const = 0;
CV_WRAP virtual float getLambda() const = 0;
CV_WRAP virtual void setLambda(float lambda) = 0;
CV_WRAP virtual int getSamples() const = 0;
CV_WRAP virtual void setSamples(int samples) = 0;
CV_WRAP virtual int getSamples() const = 0;
CV_WRAP virtual void setSamples(int samples) = 0;
};
CV_EXPORTS_W Ptr<CalibrateDebevec> createCalibrateDebevec(int samples = 50, float lambda = 10.0f);
@ -196,8 +217,8 @@ CV_EXPORTS_W Ptr<CalibrateDebevec> createCalibrateDebevec(int samples = 50, floa
class CV_EXPORTS_W ExposureMerge : public Algorithm
{
public:
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst,
const std::vector<float>& times, InputArray response) = 0;
};
// "Recovering High Dynamic Range Radiance Maps from Photographs", Debevec, Malik, 1997
@ -205,9 +226,9 @@ public:
class CV_EXPORTS_W MergeDebevec : public ExposureMerge
{
public:
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst, const std::vector<float>& times) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst, const std::vector<float>& times) = 0;
};
CV_EXPORTS_W Ptr<MergeDebevec> createMergeDebevec();
@ -217,18 +238,18 @@ CV_EXPORTS_W Ptr<MergeDebevec> createMergeDebevec();
class CV_EXPORTS_W MergeMertens : public ExposureMerge
{
public:
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst,
const std::vector<float>& times, InputArray response) = 0;
CV_WRAP virtual void process(InputArrayOfArrays src, OutputArray dst) = 0;
CV_WRAP virtual float getContrastWeight() const = 0;
CV_WRAP virtual void setContrastWeight(float contrast_weiht) = 0;
CV_WRAP virtual float getContrastWeight() const = 0;
CV_WRAP virtual void setContrastWeight(float contrast_weiht) = 0;
CV_WRAP virtual float getSaturationWeight() const = 0;
CV_WRAP virtual void setSaturationWeight(float saturation_weight) = 0;
CV_WRAP virtual float getSaturationWeight() const = 0;
CV_WRAP virtual void setSaturationWeight(float saturation_weight) = 0;
CV_WRAP virtual float getExposureWeight() const = 0;
CV_WRAP virtual void setExposureWeight(float exposure_weight) = 0;
CV_WRAP virtual float getExposureWeight() const = 0;
CV_WRAP virtual void setExposureWeight(float exposure_weight) = 0;
};
CV_EXPORTS_W Ptr<MergeMertens>

290
modules/photo/src/align.cpp
View File

@ -51,115 +51,115 @@ namespace cv
class AlignMTBImpl : public AlignMTB
{
public:
AlignMTBImpl(int max_bits, int exclude_range) :
max_bits(max_bits),
exclude_range(exclude_range),
name("AlignMTB")
{
}
void process(InputArrayOfArrays src, OutputArrayOfArrays dst,
const std::vector<float>& times, InputArray response)
{
process(src, dst);
}
AlignMTBImpl(int max_bits, int exclude_range) :
max_bits(max_bits),
exclude_range(exclude_range),
name("AlignMTB")
{
}
void process(InputArrayOfArrays src, OutputArrayOfArrays dst,
const std::vector<float>& times, InputArray response)
{
process(src, dst);
}
void process(InputArrayOfArrays _src, OutputArray _dst)
{
std::vector<Mat> src, dst;
_src.getMatVector(src);
_dst.getMatVector(dst);
void process(InputArrayOfArrays _src, OutputArray _dst)
{
std::vector<Mat> src, dst;
_src.getMatVector(src);
_dst.getMatVector(dst);
checkImageDimensions(src);
dst.resize(src.size());
checkImageDimensions(src);
dst.resize(src.size());
size_t pivot = src.size() / 2;
dst[pivot] = src[pivot];
Mat gray_base;
cvtColor(src[pivot], gray_base, COLOR_RGB2GRAY);
size_t pivot = src.size() / 2;
dst[pivot] = src[pivot];
Mat gray_base;
cvtColor(src[pivot], gray_base, COLOR_RGB2GRAY);
for(size_t i = 0; i < src.size(); i++) {
if(i == pivot) {
continue;
}
Mat gray;
cvtColor(src[i], gray, COLOR_RGB2GRAY);
Point shift;
calculateShift(gray_base, gray, shift);
shiftMat(src[i], dst[i], shift);
}
}
for(size_t i = 0; i < src.size(); i++) {
if(i == pivot) {
continue;
}
Mat gray;
cvtColor(src[i], gray, COLOR_RGB2GRAY);
Point shift;
calculateShift(gray_base, gray, shift);
shiftMat(src[i], dst[i], shift);
}
}
void calculateShift(InputArray _img0, InputArray _img1, Point& shift)
{
Mat img0 = _img0.getMat();
Mat img1 = _img1.getMat();
CV_Assert(img0.channels() == 1 && img0.type() == img1.type());
CV_Assert(img0.size() == img0.size());
void calculateShift(InputArray _img0, InputArray _img1, Point& shift)
{
Mat img0 = _img0.getMat();
Mat img1 = _img1.getMat();
CV_Assert(img0.channels() == 1 && img0.type() == img1.type());
CV_Assert(img0.size() == img0.size());
int maxlevel = static_cast<int>(log((double)max(img0.rows, img0.cols)) / log(2.0)) - 1;
maxlevel = min(maxlevel, max_bits - 1);
int maxlevel = static_cast<int>(log((double)max(img0.rows, img0.cols)) / log(2.0)) - 1;
maxlevel = min(maxlevel, max_bits - 1);
std::vector<Mat> pyr0;
std::vector<Mat> pyr1;
buildPyr(img0, pyr0, maxlevel);
buildPyr(img1, pyr1, maxlevel);
shift = Point(0, 0);
for(int level = maxlevel; level >= 0; level--) {
shift *= 2;
Mat tb1, tb2, eb1, eb2;
computeBitmaps(pyr0[level], tb1, eb1, exclude_range);
computeBitmaps(pyr1[level], tb2, eb2, exclude_range);
std::vector<Mat> pyr0;
std::vector<Mat> pyr1;
buildPyr(img0, pyr0, maxlevel);
buildPyr(img1, pyr1, maxlevel);
shift = Point(0, 0);
for(int level = maxlevel; level >= 0; level--) {
shift *= 2;
Mat tb1, tb2, eb1, eb2;
computeBitmaps(pyr0[level], tb1, eb1, exclude_range);
computeBitmaps(pyr1[level], tb2, eb2, exclude_range);
int min_err = pyr0[level].total();
Point new_shift(shift);
for(int i = -1; i <= 1; i++) {
for(int j = -1; j <= 1; j++) {
Point test_shift = shift + Point(i, j);
Mat shifted_tb2, shifted_eb2, diff;
shiftMat(tb2, shifted_tb2, test_shift);
shiftMat(eb2, shifted_eb2, test_shift);
bitwise_xor(tb1, shifted_tb2, diff);
bitwise_and(diff, eb1, diff);
bitwise_and(diff, shifted_eb2, diff);
int err = countNonZero(diff);
if(err < min_err) {
new_shift = test_shift;
min_err = err;
}
}
}
shift = new_shift;
}
}
int min_err = pyr0[level].total();
Point new_shift(shift);
for(int i = -1; i <= 1; i++) {
for(int j = -1; j <= 1; j++) {
Point test_shift = shift + Point(i, j);
Mat shifted_tb2, shifted_eb2, diff;
shiftMat(tb2, shifted_tb2, test_shift);
shiftMat(eb2, shifted_eb2, test_shift);
bitwise_xor(tb1, shifted_tb2, diff);
bitwise_and(diff, eb1, diff);
bitwise_and(diff, shifted_eb2, diff);
int err = countNonZero(diff);
if(err < min_err) {
new_shift = test_shift;
min_err = err;
}
}
}
shift = new_shift;
}
}
void shiftMat(InputArray _src, OutputArray _dst, const Point shift)
{
Mat src = _src.getMat();
_dst.create(src.size(), src.type());
Mat dst = _dst.getMat();
void shiftMat(InputArray _src, OutputArray _dst, const Point shift)
{
Mat src = _src.getMat();
_dst.create(src.size(), src.type());
Mat dst = _dst.getMat();
dst = Mat::zeros(src.size(), src.type());
int width = src.cols - abs(shift.x);
int height = src.rows - abs(shift.y);
Rect dst_rect(max(shift.x, 0), max(shift.y, 0), width, height);
Rect src_rect(max(-shift.x, 0), max(-shift.y, 0), width, height);
src(src_rect).copyTo(dst(dst_rect));
}
dst = Mat::zeros(src.size(), src.type());
int width = src.cols - abs(shift.x);
int height = src.rows - abs(shift.y);
Rect dst_rect(max(shift.x, 0), max(shift.y, 0), width, height);
Rect src_rect(max(-shift.x, 0), max(-shift.y, 0), width, height);
src(src_rect).copyTo(dst(dst_rect));
}
int getMaxBits() const { return max_bits; }
void setMaxBits(int val) { max_bits = val; }
int getMaxBits() const { return max_bits; }
void setMaxBits(int val) { max_bits = val; }
int getExcludeRange() const { return exclude_range; }
void setExcludeRange(int val) { exclude_range = val; }
int getExcludeRange() const { return exclude_range; }
void setExcludeRange(int val) { exclude_range = val; }
void write(FileStorage& fs) const
void write(FileStorage& fs) const
{
fs << "name" << name
<< "max_bits" << max_bits
<< "exclude_range" << exclude_range;
<< "max_bits" << max_bits
<< "exclude_range" << exclude_range;
}
void read(const FileNode& fn)
@ -167,69 +167,69 @@ public:
FileNode n = fn["name"];
CV_Assert(n.isString() && String(n) == name);
max_bits = fn["max_bits"];
exclude_range = fn["exclude_range"];
exclude_range = fn["exclude_range"];
}
protected:
String name;
int max_bits, exclude_range;
String name;
int max_bits, exclude_range;
void downsample(Mat& src, Mat& dst)
{
dst = Mat(src.rows / 2, src.cols / 2, CV_8UC1);
void downsample(Mat& src, Mat& dst)
{
dst = Mat(src.rows / 2, src.cols / 2, CV_8UC1);
int offset = src.cols * 2;
uchar *src_ptr = src.ptr();
uchar *dst_ptr = dst.ptr();
for(int y = 0; y < dst.rows; y ++) {
uchar *ptr = src_ptr;
for(int x = 0; x < dst.cols; x++) {
dst_ptr[0] = ptr[0];
dst_ptr++;
ptr += 2;
}
src_ptr += offset;
}
}
int offset = src.cols * 2;
uchar *src_ptr = src.ptr();
uchar *dst_ptr = dst.ptr();
for(int y = 0; y < dst.rows; y ++) {
uchar *ptr = src_ptr;
for(int x = 0; x < dst.cols; x++) {
dst_ptr[0] = ptr[0];
dst_ptr++;
ptr += 2;
}
src_ptr += offset;
}
}
void buildPyr(Mat& img, std::vector<Mat>& pyr, int maxlevel)
{
pyr.resize(maxlevel + 1);
pyr[0] = img.clone();
for(int level = 0; level < maxlevel; level++) {
downsample(pyr[level], pyr[level + 1]);
}
}
void buildPyr(Mat& img, std::vector<Mat>& pyr, int maxlevel)
{
pyr.resize(maxlevel + 1);
pyr[0] = img.clone();
for(int level = 0; level < maxlevel; level++) {
downsample(pyr[level], pyr[level + 1]);
}
}
int getMedian(Mat& img)
{
int channels = 0;
Mat hist;
int hist_size = 256;
float range[] = {0, 256} ;
const float* ranges[] = {range};
calcHist(&img, 1, &channels, Mat(), hist, 1, &hist_size, ranges);
float *ptr = hist.ptr<float>();
int median = 0, sum = 0;
int thresh = img.total() / 2;
while(sum < thresh && median < 256) {
sum += static_cast<int>(ptr[median]);
median++;
}
return median;
}
int getMedian(Mat& img)
{
int channels = 0;
Mat hist;
int hist_size = 256;
float range[] = {0, 256} ;
const float* ranges[] = {range};
calcHist(&img, 1, &channels, Mat(), hist, 1, &hist_size, ranges);
float *ptr = hist.ptr<float>();
int median = 0, sum = 0;
int thresh = img.total() / 2;
while(sum < thresh && median < 256) {
sum += static_cast<int>(ptr[median]);
median++;
}
return median;
}
void computeBitmaps(Mat& img, Mat& tb, Mat& eb, int exclude_range)
{
int median = getMedian(img);
compare(img, median, tb, CMP_GT);
compare(abs(img - median), exclude_range, eb, CMP_GT);
}
void computeBitmaps(Mat& img, Mat& tb, Mat& eb, int exclude_range)
{
int median = getMedian(img);
compare(img, median, tb, CMP_GT);
compare(abs(img - median), exclude_range, eb, CMP_GT);
}
};
CV_EXPORTS_W Ptr<AlignMTB> createAlignMTB(int max_bits, int exclude_range)
{
return new AlignMTBImpl(max_bits, exclude_range);
return new AlignMTBImpl(max_bits, exclude_range);
}
}

118
modules/photo/src/calibrate.cpp
View File

@ -47,73 +47,73 @@
namespace cv
{
class CalibrateDebevecImpl : public CalibrateDebevec
{
public:
CalibrateDebevecImpl(int samples, float lambda) :
samples(samples),
lambda(lambda),
name("CalibrateDebevec"),
w(tringleWeights())
{
}
void process(InputArrayOfArrays src, OutputArray dst, std::vector<float>& times)
{
std::vector<Mat> images;
src.getMatVector(images);
dst.create(256, images[0].channels(), CV_32F);
Mat response = dst.getMat();
CalibrateDebevecImpl(int samples, float lambda) :
samples(samples),
lambda(lambda),
name("CalibrateDebevec"),
w(tringleWeights())
{
}
void process(InputArrayOfArrays src, OutputArray dst, std::vector<float>& times)
{
std::vector<Mat> images;
src.getMatVector(images);
dst.create(256, images[0].channels(), CV_32F);
Mat response = dst.getMat();
CV_Assert(!images.empty() && images.size() == times.size());
CV_Assert(images[0].depth() == CV_8U);
checkImageDimensions(images);
CV_Assert(!images.empty() && images.size() == times.size());
CV_Assert(images[0].depth() == CV_8U);
checkImageDimensions(images);
for(int channel = 0; channel < images[0].channels(); channel++) {
Mat A = Mat::zeros(samples * images.size() + 257, 256 + samples, CV_32F);
Mat B = Mat::zeros(A.rows, 1, CV_32F);
for(int channel = 0; channel < images[0].channels(); channel++) {
Mat A = Mat::zeros(samples * images.size() + 257, 256 + samples, CV_32F);
Mat B = Mat::zeros(A.rows, 1, CV_32F);
int eq = 0;
for(int i = 0; i < samples; i++) {
int eq = 0;
for(int i = 0; i < samples; i++) {
int pos = 3 * (rand() % images[0].total()) + channel;
for(size_t j = 0; j < images.size(); j++) {
int pos = 3 * (rand() % images[0].total()) + channel;
for(size_t j = 0; j < images.size(); j++) {
int val = (images[j].ptr() + pos)[0];
A.at<float>(eq, val) = w.at<float>(val);
A.at<float>(eq, 256 + i) = -w.at<float>(val);
B.at<float>(eq, 0) = w.at<float>(val) * log(times[j]);
eq++;
}
}
A.at<float>(eq, 128) = 1;
eq++;
int val = (images[j].ptr() + pos)[0];
A.at<float>(eq, val) = w.at<float>(val);
A.at<float>(eq, 256 + i) = -w.at<float>(val);
B.at<float>(eq, 0) = w.at<float>(val) * log(times[j]);
eq++;
}
}
A.at<float>(eq, 128) = 1;
eq++;
for(int i = 0; i < 254; i++) {
A.at<float>(eq, i) = lambda * w.at<float>(i + 1);
A.at<float>(eq, i + 1) = -2 * lambda * w.at<float>(i + 1);
A.at<float>(eq, i + 2) = lambda * w.at<float>(i + 1);
eq++;
}
Mat solution;
solve(A, B, solution, DECOMP_SVD);
solution.rowRange(0, 256).copyTo(response.col(channel));
}
exp(response, response);
}
for(int i = 0; i < 254; i++) {
A.at<float>(eq, i) = lambda * w.at<float>(i + 1);
A.at<float>(eq, i + 1) = -2 * lambda * w.at<float>(i + 1);
A.at<float>(eq, i + 2) = lambda * w.at<float>(i + 1);
eq++;
}
Mat solution;
solve(A, B, solution, DECOMP_SVD);
solution.rowRange(0, 256).copyTo(response.col(channel));
}
exp(response, response);
}
int getSamples() const { return samples; }
void setSamples(int val) { samples = val; }
int getSamples() const { return samples; }
void setSamples(int val) { samples = val; }
float getLambda() const { return lambda; }
void setLambda(float val) { lambda = val; }
float getLambda() const { return lambda; }
void setLambda(float val) { lambda = val; }
void write(FileStorage& fs) const
void write(FileStorage& fs) const
{
fs << "name" << name
<< "samples" << samples
<< "lambda" << lambda;
<< "samples" << samples
<< "lambda" << lambda;
}
void read(const FileNode& fn)
@ -121,19 +121,19 @@ public:
FileNode n = fn["name"];
CV_Assert(n.isString() && String(n) == name);
samples = fn["samples"];
lambda = fn["lambda"];
lambda = fn["lambda"];
}
protected:
String name;
int samples;
float lambda;
Mat w;
String name;
int samples;
float lambda;
Mat w;
};
Ptr<CalibrateDebevec> createCalibrateDebevec(int samples, float lambda)
{
return new CalibrateDebevecImpl(samples, lambda);
return new CalibrateDebevecImpl(samples, lambda);
}
}

42
modules/photo/src/hdr_common.cpp
View File

@ -49,26 +49,38 @@ namespace cv
void checkImageDimensions(const std::vector<Mat>& images)
{
CV_Assert(!images.empty());
int width = images[0].cols;
int height = images[0].rows;
int type = images[0].type();
CV_Assert(!images.empty());
int width = images[0].cols;
int height = images[0].rows;
int type = images[0].type();
for(size_t i = 0; i < images.size(); i++) {
CV_Assert(images[i].cols == width && images[i].rows == height);
CV_Assert(images[i].type() == type);
}
for(size_t i = 0; i < images.size(); i++) {
CV_Assert(images[i].cols == width && images[i].rows == height);
CV_Assert(images[i].type() == type);
}
}
Mat tringleWeights()
{
Mat w(256, 3, CV_32F);
for(int i = 0; i < 256; i++) {
for(int j = 0; j < 3; j++) {
w.at<float>(i, j) = i < 128 ? i + 1.0f : 256.0f - i;
}
}
return w;
Mat w(256, 3, CV_32F);
for(int i = 0; i < 256; i++) {
for(int j = 0; j < 3; j++) {
w.at<float>(i, j) = i < 128 ? i + 1.0f : 256.0f - i;
}
}
return w;
}
void mapLuminance(Mat src, Mat dst, Mat lum, Mat new_lum, float saturation)
{
std::vector<Mat> channels(3);
split(src, channels);
for(int i = 0; i < 3; i++) {
channels[i] = channels[i].mul(1.0f / lum);
pow(channels[i], saturation, channels[i]);
channels[i] = channels[i].mul(new_lum);
}
merge(channels, dst);
}
};

2
modules/photo/src/hdr_common.hpp
View File

@ -53,6 +53,8 @@ void checkImageDimensions(const std::vector<Mat>& images);
Mat tringleWeights();
void mapLuminance(Mat src, Mat dst, Mat lum, Mat new_lum, float saturation);
};
#endif

322
modules/photo/src/merge.cpp
View File

@ -52,193 +52,193 @@ namespace cv
class MergeDebevecImpl : public MergeDebevec
{
public:
MergeDebevecImpl() :
name("MergeDebevec"),
weights(tringleWeights())
{
}
void process(InputArrayOfArrays src, OutputArray dst, const std::vector<float>& times, InputArray input_response)
{
std::vector<Mat> images;
src.getMatVector(images);
dst.create(images[0].size(), CV_MAKETYPE(CV_32F, images[0].channels()));
Mat result = dst.getMat();
MergeDebevecImpl() :
name("MergeDebevec"),
weights(tringleWeights())
{
}
void process(InputArrayOfArrays src, OutputArray dst, const std::vector<float>& times, InputArray input_response)
{
std::vector<Mat> images;
src.getMatVector(images);
dst.create(images[0].size(), CV_MAKETYPE(CV_32F, images[0].channels()));
Mat result = dst.getMat();
CV_Assert(images.size() == times.size());
CV_Assert(images[0].depth() == CV_8U);
checkImageDimensions(images);
CV_Assert(images.size() == times.size());
CV_Assert(images[0].depth() == CV_8U);
checkImageDimensions(images);
Mat response = input_response.getMat();
CV_Assert(response.rows == 256 && response.cols >= images[0].channels());
Mat log_response;
log(response, log_response);
std::vector<float> exp_times(times.size());
for(size_t i = 0; i < exp_times.size(); i++) {
exp_times[i] = logf(times[i]);
}
int channels = images[0].channels();
float *res_ptr = result.ptr<float>();
for(size_t pos = 0; pos < result.total(); pos++, res_ptr += channels) {
Mat response = input_response.getMat();
CV_Assert(response.rows == 256 && response.cols >= images[0].channels());
Mat log_response;
log(response, log_response);
std::vector<float> exp_times(times.size());
for(size_t i = 0; i < exp_times.size(); i++) {
exp_times[i] = logf(times[i]);
}
int channels = images[0].channels();
float *res_ptr = result.ptr<float>();
for(size_t pos = 0; pos < result.total(); pos++, res_ptr += channels) {
std::vector<float> sum(channels, 0);
float weight_sum = 0;
for(size_t im = 0; im < images.size(); im++) {
std::vector<float> sum(channels, 0);
float weight_sum = 0;
for(size_t im = 0; im < images.size(); im++) {
uchar *img_ptr = images[im].ptr() + channels * pos;
float w = 0;
for(int channel = 0; channel < channels; channel++) {
w += weights.at<float>(img_ptr[channel]);
}
w /= channels;
weight_sum += w;
for(int channel = 0; channel < channels; channel++) {
sum[channel] += w * (log_response.at<float>(img_ptr[channel], channel) - exp_times[im]);
}
}
for(int channel = 0; channel < channels; channel++) {
res_ptr[channel] = exp(sum[channel] / weight_sum);
}
}
}
uchar *img_ptr = images[im].ptr() + channels * pos;
float w = 0;
for(int channel = 0; channel < channels; channel++) {
w += weights.at<float>(img_ptr[channel]);
}
w /= channels;
weight_sum += w;
for(int channel = 0; channel < channels; channel++) {
sum[channel] += w * (log_response.at<float>(img_ptr[channel], channel) - exp_times[im]);
}
}
for(int channel = 0; channel < channels; channel++) {
res_ptr[channel] = exp(sum[channel] / weight_sum);
}
}
}
void process(InputArrayOfArrays src, OutputArray dst, const std::vector<float>& times)
{
Mat response(256, 3, CV_32F);
for(int i = 0; i < 256; i++) {
for(int j = 0; j < 3; j++) {
response.at<float>(i, j) = max(i, 1);
}
}
process(src, dst, times, response);
}
void process(InputArrayOfArrays src, OutputArray dst, const std::vector<float>& times)
{
Mat response(256, 3, CV_32F);
for(int i = 0; i < 256; i++) {
for(int j = 0; j < 3; j++) {
response.at<float>(i, j) = static_cast<float>(max(i, 1));
}
}
process(src, dst, times, response);
}
protected:
String name;
Mat weights;
String name;
Mat weights;
};
Ptr<MergeDebevec> createMergeDebevec()
{
return new MergeDebevecImpl;
return new MergeDebevecImpl;
}
class MergeMertensImpl : public MergeMertens
{
public:
MergeMertensImpl(float wcon, float wsat, float wexp) :
wcon(wcon),
wsat(wsat),
wexp(wexp),
name("MergeMertens")
{
}
void process(InputArrayOfArrays src, OutputArrayOfArrays dst, const std::vector<float>& times, InputArray response)
{
process(src, dst);
}
MergeMertensImpl(float wcon, float wsat, float wexp) :
wcon(wcon),
wsat(wsat),
wexp(wexp),
name("MergeMertens")
{
}
void process(InputArrayOfArrays src, OutputArrayOfArrays dst, const std::vector<float>& times, InputArray response)
{
process(src, dst);
}
void process(InputArrayOfArrays src, OutputArray dst)
{
std::vector<Mat> images;
src.getMatVector(images);
checkImageDimensions(images);
void process(InputArrayOfArrays src, OutputArray dst)
{
std::vector<Mat> images;
src.getMatVector(images);
checkImageDimensions(images);
std::vector<Mat> weights(images.size());
Mat weight_sum = Mat::zeros(images[0].size(), CV_32FC1);
for(size_t im = 0; im < images.size(); im++) {
Mat img, gray, contrast, saturation, wellexp;
std::vector<Mat> channels(3);
std::vector<Mat> weights(images.size());
Mat weight_sum = Mat::zeros(images[0].size(), CV_32FC1);
for(size_t im = 0; im < images.size(); im++) {
Mat img, gray, contrast, saturation, wellexp;
std::vector<Mat> channels(3);
images[im].convertTo(img, CV_32FC3, 1.0/255.0);
cvtColor(img, gray, COLOR_RGB2GRAY);
split(img, channels);
images[im].convertTo(img, CV_32FC3, 1.0/255.0);
cvtColor(img, gray, COLOR_RGB2GRAY);
split(img, channels);
Laplacian(gray, contrast, CV_32F);
contrast = abs(contrast);
Laplacian(gray, contrast, CV_32F);
contrast = abs(contrast);
Mat mean = (channels[0] + channels[1] + channels[2]) / 3.0f;
saturation = Mat::zeros(channels[0].size(), CV_32FC1);
for(int i = 0; i < 3; i++) {
Mat deviation = channels[i] - mean;
pow(deviation, 2.0, deviation);
saturation += deviation;
}
sqrt(saturation, saturation);
Mat mean = (channels[0] + channels[1] + channels[2]) / 3.0f;
saturation = Mat::zeros(channels[0].size(), CV_32FC1);
for(int i = 0; i < 3; i++) {
Mat deviation = channels[i] - mean;
pow(deviation, 2.0, deviation);
saturation += deviation;
}
sqrt(saturation, saturation);
wellexp = Mat::ones(gray.size(), CV_32FC1);
for(int i = 0; i < 3; i++) {
Mat exp = channels[i] - 0.5f;
pow(exp, 2, exp);
exp = -exp / 0.08;
wellexp = wellexp.mul(exp);
}
wellexp = Mat::ones(gray.size(), CV_32FC1);
for(int i = 0; i < 3; i++) {
Mat exp = channels[i] - 0.5f;
pow(exp, 2, exp);
exp = -exp / 0.08;
wellexp = wellexp.mul(exp);
}
pow(contrast, wcon, contrast);
pow(saturation, wsat, saturation);
pow(wellexp, wexp, wellexp);
pow(contrast, wcon, contrast);
pow(saturation, wsat, saturation);
pow(wellexp, wexp, wellexp);
weights[im] = contrast;
weights[im] = weights[im].mul(saturation);
weights[im] = weights[im].mul(wellexp);
weight_sum += weights[im];
}
int maxlevel = static_cast<int>(logf(static_cast<float>(max(images[0].rows, images[0].cols))) / logf(2.0)) - 1;
std::vector<Mat> res_pyr(maxlevel + 1);
weights[im] = contrast;
weights[im] = weights[im].mul(saturation);
weights[im] = weights[im].mul(wellexp);
weight_sum += weights[im];
}
int maxlevel = static_cast<int>(logf(static_cast<float>(max(images[0].rows, images[0].cols))) / logf(2.0)) - 1;
std::vector<Mat> res_pyr(maxlevel + 1);
for(size_t im = 0; im < images.size(); im++) {
weights[im] /= weight_sum;
Mat img;
images[im].convertTo(img, CV_32FC3, 1/255.0);
std::vector<Mat> img_pyr, weight_pyr;
buildPyramid(img, img_pyr, maxlevel);
buildPyramid(weights[im], weight_pyr, maxlevel);
for(int lvl = 0; lvl < maxlevel; lvl++) {
Mat up;
pyrUp(img_pyr[lvl + 1], up, img_pyr[lvl].size());
img_pyr[lvl] -= up;
}
for(int lvl = 0; lvl <= maxlevel; lvl++) {
std::vector<Mat> channels(3);
split(img_pyr[lvl], channels);
for(int i = 0; i < 3; i++) {
channels[i] = channels[i].mul(weight_pyr[lvl]);
}
merge(channels, img_pyr[lvl]);
if(res_pyr[lvl].empty()) {
res_pyr[lvl] = img_pyr[lvl];
} else {
res_pyr[lvl] += img_pyr[lvl];
}
}
}
for(int lvl = maxlevel; lvl > 0; lvl--) {
Mat up;
pyrUp(res_pyr[lvl], up, res_pyr[lvl - 1].size());
res_pyr[lvl - 1] += up;
}
dst.create(images[0].size(), CV_32FC3);
res_pyr[0].copyTo(dst.getMat());
}
for(size_t im = 0; im < images.size(); im++) {
weights[im] /= weight_sum;
Mat img;
images[im].convertTo(img, CV_32FC3, 1/255.0);
std::vector<Mat> img_pyr, weight_pyr;
buildPyramid(img, img_pyr, maxlevel);
buildPyramid(weights[im], weight_pyr, maxlevel);
for(int lvl = 0; lvl < maxlevel; lvl++) {
Mat up;
pyrUp(img_pyr[lvl + 1], up, img_pyr[lvl].size());
img_pyr[lvl] -= up;
}
for(int lvl = 0; lvl <= maxlevel; lvl++) {
std::vector<Mat> channels(3);
split(img_pyr[lvl], channels);
for(int i = 0; i < 3; i++) {
channels[i] = channels[i].mul(weight_pyr[lvl]);
}
merge(channels, img_pyr[lvl]);
if(res_pyr[lvl].empty()) {
res_pyr[lvl] = img_pyr[lvl];
} else {
res_pyr[lvl] += img_pyr[lvl];
}
}
}
for(int lvl = maxlevel; lvl > 0; lvl--) {
Mat up;
pyrUp(res_pyr[lvl], up, res_pyr[lvl - 1].size());
res_pyr[lvl - 1] += up;
}
dst.create(images[0].size(), CV_32FC3);
res_pyr[0].copyTo(dst.getMat());
}
float getContrastWeight() const { return wcon; }
void setContrastWeight(float val) { wcon = val; }
float getContrastWeight() const { return wcon; }
void setContrastWeight(float val) { wcon = val; }
float getSaturationWeight() const { return wsat; }
void setSaturationWeight(float val) { wsat = val; }
float getSaturationWeight() const { return wsat; }
void setSaturationWeight(float val) { wsat = val; }
float getExposureWeight() const { return wexp; }
void setExposureWeight(float val) { wexp = val; }
float getExposureWeight() const { return wexp; }
void setExposureWeight(float val) { wexp = val; }
void write(FileStorage& fs) const
void write(FileStorage& fs) const
{
fs << "name" << name
<< "contrast_weight" << wcon
<< "saturation_weight" << wsat
<< "exposure_weight" << wexp;
<< "contrast_weight" << wcon
<< "saturation_weight" << wsat
<< "exposure_weight" << wexp;
}
void read(const FileNode& fn)
@ -246,18 +246,18 @@ public:
FileNode n = fn["name"];
CV_Assert(n.isString() && String(n) == name);
wcon = fn["contrast_weight"];
wsat = fn["saturation_weight"];
wexp = fn["exposure_weight"];
wsat = fn["saturation_weight"];
wexp = fn["exposure_weight"];
}
protected:
String name;
float wcon, wsat, wexp;
String name;
float wcon, wsat, wexp;
};
Ptr<MergeMertens> createMergeMertens(float wcon, float wsat, float wexp)
{
return new MergeMertensImpl(wcon, wsat, wexp);
return new MergeMertensImpl(wcon, wsat, wexp);
}
}

558
modules/photo/src/tonemap.cpp
View File

@ -43,6 +43,7 @@
#include "precomp.hpp"
#include "opencv2/photo.hpp"
#include "opencv2/imgproc.hpp"
#include "hdr_common.hpp"
namespace cv
{
@ -50,32 +51,32 @@ namespace cv
class TonemapLinearImpl : public TonemapLinear
{
public:
TonemapLinearImpl(float gamma) : gamma(gamma), name("TonemapLinear")
{
}
TonemapLinearImpl(float gamma) : gamma(gamma), name("TonemapLinear")
{
}
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat dst = _dst.getMat();
double min, max;
minMaxLoc(src, &min, &max);
if(max - min > DBL_EPSILON) {
dst = (src - min) / (max - min);
} else {
src.copyTo(dst);
}
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat dst = _dst.getMat();
double min, max;
minMaxLoc(src, &min, &max);
if(max - min > DBL_EPSILON) {
dst = (src - min) / (max - min);
} else {
src.copyTo(dst);
}
pow(dst, 1.0f / gamma, dst);
}
pow(dst, 1.0f / gamma, dst);
}
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
void write(FileStorage& fs) const
void write(FileStorage& fs) const
{
fs << "name" << name
<< "gamma" << gamma;
@ -89,79 +90,76 @@ public:
}
protected:
String name;
float gamma;
String name;
float gamma;
};
Ptr<TonemapLinear> createTonemapLinear(float gamma)
{
return new TonemapLinearImpl(gamma);
return new TonemapLinearImpl(gamma);
}
class TonemapDragoImpl : public TonemapDrago
{
public:
TonemapDragoImpl(float gamma, float bias) :
gamma(gamma),
TonemapDragoImpl(float gamma, float saturation, float bias) :
gamma(gamma),
saturation(saturation),
bias(bias),
name("TonemapLinear")
{
}
name("TonemapDrago")
{
}
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat img = _dst.getMat();
Ptr<TonemapLinear> linear = createTonemapLinear(1.0f);
linear->process(src, img);
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat img = _dst.getMat();
Ptr<TonemapLinear> linear = createTonemapLinear(1.0f);
linear->process(src, img);
Mat gray_img;
cvtColor(img, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
float mean = expf(static_cast<float>(sum(log_img)[0]) / log_img.total());
gray_img /= mean;
log_img.release();
Mat gray_img;
cvtColor(img, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
float mean = expf(static_cast<float>(sum(log_img)[0]) / log_img.total());
gray_img /= mean;
log_img.release();
double max;
minMaxLoc(gray_img, NULL, &max);
double max;
minMaxLoc(gray_img, NULL, &max);
Mat map;
log(gray_img + 1.0f, map);
Mat div;
pow(gray_img / static_cast<float>(max), logf(bias) / logf(0.5f), div);
log(2.0f + 8.0f * div, div);
map = map.mul(1.0f / div);
map = map.mul(1.0f / gray_img);
div.release();
gray_img.release();
Mat map;
log(gray_img + 1.0f, map);
Mat div;
pow(gray_img / static_cast<float>(max), logf(bias) / logf(0.5f), div);
log(2.0f + 8.0f * div, div);
map = map.mul(1.0f / div);
div.release();
std::vector<Mat> channels(3);
split(img, channels);
for(int i = 0; i < 3; i++) {
channels[i] = channels[i].mul(map);
}
map.release();
merge(channels, img);
linear->setGamma(gamma);
linear->process(img, img);
}
mapLuminance(img, img, gray_img, map, saturation);
linear->setGamma(gamma);
linear->process(img, img);
}
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getBias() const { return bias; }
void setBias(float val) { bias = val; }
float getSaturation() const { return saturation; }
void setSaturation(float val) { saturation = val; }
void write(FileStorage& fs) const
float getBias() const { return bias; }
void setBias(float val) { bias = val; }
void write(FileStorage& fs) const
{
fs << "name" << name
<< "gamma" << gamma
<< "bias" << bias;
<< "bias" << bias
<< "saturation" << saturation;
}
void read(const FileNode& fn)
@ -169,82 +167,82 @@ public:
FileNode n = fn["name"];
CV_Assert(n.isString() && String(n) == name);
gamma = fn["gamma"];
bias = fn["bias"];
bias = fn["bias"];
saturation = fn["saturation"];
}
protected:
String name;
float gamma, bias;
String name;
float gamma, saturation, bias;
};
Ptr<TonemapDrago> createTonemapDrago(float gamma, float bias)
Ptr<TonemapDrago> createTonemapDrago(float gamma, float saturation, float bias)
{
return new TonemapDragoImpl(gamma, bias);
return new TonemapDragoImpl(gamma, saturation, bias);
}
class TonemapDurandImpl : public TonemapDurand
{
public:
TonemapDurandImpl(float gamma, float contrast, float sigma_color, float sigma_space) :
gamma(gamma),
TonemapDurandImpl(float gamma, float saturation, float contrast, float sigma_color, float sigma_space) :
gamma(gamma),
saturation(saturation),
contrast(contrast),
sigma_color(sigma_color),
sigma_space(sigma_space),
name("TonemapDurand")
{
}
sigma_color(sigma_color),
sigma_space(sigma_space),
name("TonemapDurand")
{
}
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat dst = _dst.getMat();
Mat gray_img;
cvtColor(src, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
Mat map_img;
bilateralFilter(log_img, map_img, -1, sigma_color, sigma_space);
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat img = _dst.getMat();
Ptr<TonemapLinear> linear = createTonemapLinear(1.0f);
linear->process(src, img);
Mat gray_img;
cvtColor(img, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
Mat map_img;
bilateralFilter(log_img, map_img, -1, sigma_color, sigma_space);
double min, max;
minMaxLoc(map_img, &min, &max);
float scale = contrast / static_cast<float>(max - min);
double min, max;
minMaxLoc(map_img, &min, &max);
float scale = contrast / static_cast<float>(max - min);
exp(map_img * (scale - 1.0f) + log_img, map_img);
log_img.release();
exp(map_img * (scale - 1.0f) + log_img, map_img);
log_img.release();
map_img = map_img.mul(1.0f / gray_img);
gray_img.release();
mapLuminance(img, img, gray_img, map_img, saturation);
pow(img, 1.0f / gamma, img);
}
std::vector<Mat> channels(3);
split(src, channels);
for(int i = 0; i < 3; i++) {
channels[i] = channels[i].mul(map_img);
}
merge(channels, dst);
pow(dst, 1.0f / gamma, dst);
}
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getSaturation() const { return saturation; }
void setSaturation(float val) { saturation = val; }
float getContrast() const { return contrast; }
void setContrast(float val) { contrast = val; }
float getContrast() const { return contrast; }
void setContrast(float val) { contrast = val; }
float getSigmaColor() const { return sigma_color; }
void setSigmaColor(float val) { sigma_color = val; }
float getSigmaColor() const { return sigma_color; }
void setSigmaColor(float val) { sigma_color = val; }
float getSigmaSpace() const { return sigma_space; }
void setSigmaSpace(float val) { sigma_space = val; }
float getSigmaSpace() const { return sigma_space; }
void setSigmaSpace(float val) { sigma_space = val; }
void write(FileStorage& fs) const
void write(FileStorage& fs) const
{
fs << "name" << name
<< "gamma" << gamma
<< "contrast" << contrast
<< "sigma_color" << sigma_color
<< "sigma_space" << sigma_space;
<< "contrast" << contrast
<< "sigma_color" << sigma_color
<< "sigma_space" << sigma_space
<< "saturation" << saturation;
}
void read(const FileNode& fn)
@ -252,95 +250,95 @@ public:
FileNode n = fn["name"];
CV_Assert(n.isString() && String(n) == name);
gamma = fn["gamma"];
contrast = fn["contrast"];
sigma_color = fn["sigma_color"];
sigma_space = fn["sigma_space"];
contrast = fn["contrast"];
sigma_color = fn["sigma_color"];
sigma_space = fn["sigma_space"];
saturation = fn["saturation"];
}
protected:
String name;
float gamma, contrast, sigma_color, sigma_space;
String name;
float gamma, saturation, contrast, sigma_color, sigma_space;
};
Ptr<TonemapDurand> createTonemapDurand(float gamma, float contrast, float sigma_color, float sigma_space)
Ptr<TonemapDurand> createTonemapDurand(float gamma, float saturation, float contrast, float sigma_color, float sigma_space)
{
return new TonemapDurandImpl(gamma, contrast, sigma_color, sigma_space);
return new TonemapDurandImpl(gamma, saturation, contrast, sigma_color, sigma_space);
}
class TonemapReinhardDevlinImpl : public TonemapReinhardDevlin
{
public:
TonemapReinhardDevlinImpl(float gamma, float intensity, float light_adapt, float color_adapt) :
gamma(gamma),
TonemapReinhardDevlinImpl(float gamma, float intensity, float light_adapt, float color_adapt) :
gamma(gamma),
intensity(intensity),
light_adapt(light_adapt),
color_adapt(color_adapt),
name("TonemapReinhardDevlin")
{
}
light_adapt(light_adapt),
color_adapt(color_adapt),
name("TonemapReinhardDevlin")
{
}
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat img = _dst.getMat();
Ptr<TonemapLinear> linear = createTonemapLinear(1.0f);
linear->process(src, img);
Mat gray_img;
cvtColor(img, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat img = _dst.getMat();
Ptr<TonemapLinear> linear = createTonemapLinear(1.0f);
linear->process(src, img);
Mat gray_img;
cvtColor(img, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
float log_mean = static_cast<float>(sum(log_img)[0] / log_img.total());
double log_min, log_max;
minMaxLoc(log_img, &log_min, &log_max);
log_img.release();
float log_mean = static_cast<float>(sum(log_img)[0] / log_img.total());
double log_min, log_max;
minMaxLoc(log_img, &log_min, &log_max);
log_img.release();
double key = static_cast<float>((log_max - log_mean) / (log_max - log_min));
float map_key = 0.3f + 0.7f * pow(static_cast<float>(key), 1.4f);
intensity = exp(-intensity);
Scalar chan_mean = mean(img);
float gray_mean = static_cast<float>(mean(gray_img)[0]);
double key = static_cast<float>((log_max - log_mean) / (log_max - log_min));
float map_key = 0.3f + 0.7f * pow(static_cast<float>(key), 1.4f);
intensity = exp(-intensity);
Scalar chan_mean = mean(img);
float gray_mean = static_cast<float>(mean(gray_img)[0]);
std::vector<Mat> channels(3);
split(img, channels);
std::vector<Mat> channels(3);
split(img, channels);
for(int i = 0; i < 3; i++) {
float global = color_adapt * static_cast<float>(chan_mean[i]) + (1.0f - color_adapt) * gray_mean;
Mat adapt = color_adapt * channels[i] + (1.0f - color_adapt) * gray_img;
adapt = light_adapt * adapt + (1.0f - light_adapt) * global;
pow(intensity * adapt, map_key, adapt);
channels[i] = channels[i].mul(1.0f / (adapt + channels[i]));
}
gray_img.release();
merge(channels, img);
linear->setGamma(gamma);
linear->process(img, img);
}
for(int i = 0; i < 3; i++) {
float global = color_adapt * static_cast<float>(chan_mean[i]) + (1.0f - color_adapt) * gray_mean;
Mat adapt = color_adapt * channels[i] + (1.0f - color_adapt) * gray_img;
adapt = light_adapt * adapt + (1.0f - light_adapt) * global;
pow(intensity * adapt, map_key, adapt);
channels[i] = channels[i].mul(1.0f / (adapt + channels[i]));
}
gray_img.release();
merge(channels, img);
linear->setGamma(gamma);
linear->process(img, img);
}
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getIntensity() const { return intensity; }
void setIntensity(float val) { intensity = val; }
float getIntensity() const { return intensity; }
void setIntensity(float val) { intensity = val; }
float getLightAdaptation() const { return light_adapt; }
void setLightAdaptation(float val) { light_adapt = val; }
float getLightAdaptation() const { return light_adapt; }
void setLightAdaptation(float val) { light_adapt = val; }
float getColorAdaptation() const { return color_adapt; }
void setColorAdaptation(float val) { color_adapt = val; }
float getColorAdaptation() const { return color_adapt; }
void setColorAdaptation(float val) { color_adapt = val; }
void write(FileStorage& fs) const
void write(FileStorage& fs) const
{
fs << "name" << name
<< "gamma" << gamma
<< "intensity" << intensity
<< "light_adapt" << light_adapt
<< "color_adapt" << color_adapt;
<< "intensity" << intensity
<< "light_adapt" << light_adapt
<< "color_adapt" << color_adapt;
}
void read(const FileNode& fn)
@ -348,19 +346,187 @@ public:
FileNode n = fn["name"];
CV_Assert(n.isString() && String(n) == name);
gamma = fn["gamma"];
intensity = fn["intensity"];
light_adapt = fn["light_adapt"];
color_adapt = fn["color_adapt"];
intensity = fn["intensity"];
light_adapt = fn["light_adapt"];
color_adapt = fn["color_adapt"];
}
protected:
String name;
float gamma, intensity, light_adapt, color_adapt;
String name;
float gamma, intensity, light_adapt, color_adapt;
};
Ptr<TonemapReinhardDevlin> createTonemapReinhardDevlin(float gamma, float contrast, float sigma_color, float sigma_space)
{
return new TonemapReinhardDevlinImpl(gamma, contrast, sigma_color, sigma_space);
return new TonemapReinhardDevlinImpl(gamma, contrast, sigma_color, sigma_space);
}
class TonemapMantiukImpl : public TonemapMantiuk
{
public:
TonemapMantiukImpl(float gamma, float scale, float saturation) :
gamma(gamma),
scale(scale),
saturation(saturation),
name("TonemapMantiuk")
{
}
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat img = _dst.getMat();
Ptr<TonemapLinear> linear = createTonemapLinear(1.0f);
linear->process(src, img);
Mat gray_img;
cvtColor(img, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
std::vector<Mat> x_contrast, y_contrast;
getContrast(log_img, x_contrast, y_contrast);
for(size_t i = 0; i < x_contrast.size(); i++) {
mapContrast(x_contrast[i], scale);
mapContrast(y_contrast[i], scale);
}
Mat right(src.size(), CV_32F);
calculateSum(x_contrast, y_contrast, right);
Mat p, r, product, x = log_img;
calculateProduct(x, r);
r = right - r;
r.copyTo(p);
const float target_error = 1e-3f;
float target_norm = static_cast<float>(right.dot(right)) * powf(target_error, 2.0f);
int max_iterations = 100;
float rr = static_cast<float>(r.dot(r));
for(int i = 0; i < max_iterations; i++)
{
calculateProduct(p, product);
float alpha = rr / static_cast<float>(p.dot(product));
r -= alpha * product;
x += alpha * p;
float new_rr = static_cast<float>(r.dot(r));
p = r + (new_rr / rr) * p;
rr = new_rr;
if(rr < target_norm) {
break;
}
}
exp(x, x);
mapLuminance(img, img, gray_img, x, saturation);
linear = createTonemapLinear(gamma);
linear->process(img, img);
}
float getGamma() const { return gamma; }
void setGamma(float val) { gamma = val; }
float getScale() const { return scale; }
void setScale(float val) { scale = val; }
float getSaturation() const { return saturation; }
void setSaturation(float val) { saturation = val; }
void write(FileStorage& fs) const
{
fs << "name" << name
<< "gamma" << gamma
<< "scale" << scale
<< "saturation" << saturation;
}
void read(const FileNode& fn)
{
FileNode n = fn["name"];
CV_Assert(n.isString() && String(n) == name);
gamma = fn["gamma"];
scale = fn["scale"];
saturation = fn["saturation"];
}
protected:
String name;
float gamma, scale, saturation;
void signedPow(Mat src, float power, Mat& dst)
{
Mat sign = (src > 0);
sign.convertTo(sign, CV_32F, 1/255.0f);
sign = sign * 2 - 1;
pow(abs(src), power, dst);
dst = dst.mul(sign);
}
void mapContrast(Mat& contrast, float scale)
{
const float response_power = 0.4185f;
signedPow(contrast, response_power, contrast);
contrast *= scale;
signedPow(contrast, 1.0f / response_power, contrast);
}
void getGradient(Mat src, Mat& dst, int pos)
{
dst = Mat::zeros(src.size(), CV_32F);
Mat a, b;
Mat grad = src.colRange(1, src.cols) - src.colRange(0, src.cols - 1);
grad.copyTo(dst.colRange(pos, src.cols + pos - 1));
if(pos == 1) {
src.col(0).copyTo(dst.col(0));
}
}
void getContrast(Mat src, std::vector<Mat>& x_contrast, std::vector<Mat>& y_contrast)
{
int levels = static_cast<int>(logf(static_cast<float>(min(src.rows, src.cols))) / logf(2.0f));
x_contrast.resize(levels);
y_contrast.resize(levels);
Mat layer;
src.copyTo(layer);
for(int i = 0; i < levels; i++) {
getGradient(layer, x_contrast[i], 0);
getGradient(layer.t(), y_contrast[i], 0);
resize(layer, layer, Size(layer.cols / 2, layer.rows / 2));
}
}
void calculateSum(std::vector<Mat>& x_contrast, std::vector<Mat>& y_contrast, Mat& sum)
{
sum = Mat::zeros(x_contrast[x_contrast.size() - 1].size(), CV_32F);
for(int i = x_contrast.size() - 1; i >= 0; i--)
{
Mat grad_x, grad_y;
getGradient(x_contrast[i], grad_x, 1);
getGradient(y_contrast[i], grad_y, 1);
resize(sum, sum, x_contrast[i].size());
sum += grad_x + grad_y.t();
}
}
void calculateProduct(Mat src, Mat& dst)
{
std::vector<Mat> x_contrast, y_contrast;
getContrast(src, x_contrast, y_contrast);
calculateSum(x_contrast, y_contrast, dst);
}
};
Ptr<TonemapMantiuk> createTonemapMantiuk(float gamma, float scale, float saturation)
{
return new TonemapMantiukImpl(gamma, scale, saturation);
}
}

11
modules/photo/test/test_hdr.cpp
View File

@ -91,12 +91,11 @@ void loadResponseCSV(String path, Mat& response)
TEST(Photo_Tonemap, regression)
{
string test_path = string(cvtest::TS::ptr()->get_data_path()) + "hdr/";
string test_path = string(cvtest::TS::ptr()->get_data_path()) + "hdr/tonemap/";
Mat img, expected, result;
loadImage(test_path + "rle.hdr", img);
loadImage(test_path + "image.hdr", img);
float gamma = 2.2f;
test_path += "tonemap/";
Ptr<TonemapLinear> linear = createTonemapLinear(gamma);
linear->process(img, result);
@ -121,6 +120,12 @@ TEST(Photo_Tonemap, regression)
loadImage(test_path + "reinharddevlin.png", expected);
result.convertTo(result, CV_8UC3, 255);
checkEqual(result, expected, 0);
Ptr<TonemapMantiuk> mantiuk = createTonemapMantiuk(gamma);
mantiuk->process(img, result);
loadImage(test_path + "mantiuk.png", expected);
result.convertTo(result, CV_8UC3, 255);
checkEqual(result, expected, 0);
}
TEST(Photo_AlignMTB, regression)