CLIFIntegralResult
clifGrayscaleIntegral(const IplImage* source,
CLIFEnvironmentData* data,
const cl_bool use_opencl)
{
CLIFIntegralResult ret;
if(!use_opencl) {
IplImage* grayscale = cvCreateImage(cvSize(source->width, source->height), IPL_DEPTH_8U, 1);
cvCvtColor(source, grayscale, CV_BGR2GRAY);
ret.image = cvCreateMat(source->height + 1, source->width + 1, CV_32SC1);
ret.square_image = cvCreateMat(source->height + 1, source->width + 1, CV_64FC1);
cvIntegral(grayscale, ret.image, ret.square_image);
cvReleaseImage(&grayscale);
return ret;
}
cl_int error = CL_SUCCESS;
// Init buffer
error = clEnqueueWriteBuffer(data->environment.queue, data->bgr_to_gray_data.buffers[0], CL_FALSE, 0, source->widthStep * source->height, source->imageData, 0, NULL, NULL);
clCheckOrExit(error);
// Run kernel
error = clEnqueueNDRangeKernel(data->environment.queue, data->environment.kernels[0], 2, NULL, data->bgr_to_gray_data.global_size, data->bgr_to_gray_data.local_size, 0, NULL, NULL);
clCheckOrExit(error);
// Set as arg the output of greyscale
clSetKernelArg(data->environment.kernels[1], 0, sizeof(cl_mem), &(data->bgr_to_gray_data.buffers[1]));
clCheckOrExit(error);
// Run sum rows kernel
error = clEnqueueNDRangeKernel(data->environment.queue, data->environment.kernels[1], 1, NULL, &(data->integral_image_data.global_size[0]), &(data->integral_image_data.local_size[0]), 0, NULL, NULL);
clCheckOrExit(error);
// Run sum cols kernel
error = clEnqueueNDRangeKernel(data->environment.queue, data->environment.kernels[2], 1, NULL, &(data->integral_image_data.global_size[1]), &(data->integral_image_data.local_size[1]), 0, NULL, NULL);
clCheckOrExit(error);
// Read result
cl_uint* result = (cl_uint*)clEnqueueMapBuffer(data->environment.queue, data->integral_image_data.buffers[3], CL_TRUE, CL_MAP_READ, 0, (source->width + 1) * (source->height + 1) * sizeof(cl_uint), 0, NULL, NULL, &error);
clCheckOrExit(error);
cl_ulong* square_result = (cl_ulong*)clEnqueueMapBuffer(data->environment.queue, data->integral_image_data.buffers[4], CL_TRUE, CL_MAP_READ, 0, (source->width + 1) * (source->height + 1) * sizeof(cl_ulong), 0, NULL, NULL, &error);
clCheckOrExit(error);
data->integral_image_data.ptr = result;
// Return
ret.image = cvCreateMatHeader(source->height + 1, source->width + 1, CV_32SC1);
cvSetData(ret.image, result, source->width + 1);
ret.square_image = cvCreateMatHeader(source->height + 1, source->width + 1, CV_64FC1);
cvSetData(ret.square_image, square_result, source->width + 1);
return ret;
}
void extractHaarFeatures(const IplImage* img, Mat& haar)
{
CvSize size = cvSize(IMAGE_RESIZE, IMAGE_RESIZE);
CvSize size2 = cvSize(INTEGRAL_SIZE, INTEGRAL_SIZE);
CvSize img_size = cvGetSize(img);
IplImage* ipl= cvCreateImage(img_size,8,0);
if(img->nChannels==3)
{
cvCvtColor(img,ipl,CV_BGR2GRAY);
}
else
{
cvCopy(img,ipl,0);
}
if((size.width!=img_size.width)|| (size.height!=img_size.height))
{
IplImage* tmpsize=cvCreateImage(size,IPL_DEPTH_8U,0);
cvResize(ipl,tmpsize,CV_INTER_LINEAR);
cvReleaseImage( &ipl);
ipl=cvCreateImage(size,IPL_DEPTH_8U,0);
cvCopy(tmpsize,ipl,0);
cvReleaseImage( &tmpsize);
}
IplImage* temp = cvCreateImage(size, IPL_DEPTH_64F, 0);
cvCvtScale(ipl,temp);
cvNormalize(temp, temp, 0, 1, CV_MINMAX);
haar.release();
haar = Mat::zeros(1, NUM_HAAR_FEATURES, CV_32FC1);
IplImage* integral=cvCreateImage(size2,IPL_DEPTH_64F,0);
CvMat * sqSum = cvCreateMat(temp->height + 1, temp->width + 1, CV_64FC1);
cvIntegral(temp, integral, sqSum);
cvReleaseMat(&sqSum);
int actualSize = 0;
// top left
for(int i = 0; i < 100; i+= 10) {
for(int j = 0; j < 100; j+= 10) {
// bottom right
for(int m = i+10; m<=100; m+=10) {
for(int n = j+10; n<=100; n+=10) {
haar.at<float>(0, actualSize++) = getIntegralRectValue(integral, i, j, m, n);
}
}
}
}
cvReleaseImage(&ipl);
cvReleaseImage(&temp);
cvReleaseImage(&integral);
}
void getBriefDescriptors2(int descriptors[getNumberOfKeypoints()][DESC_LEN],
const CvPoint kpts[getNumberOfKeypoints()], IplImage* img) {
unsigned int i;
// Make sure that input image contains only one color channel:
assert(img->nChannels == 1);
// Check whether all the keypoints are inside the subimage:
CvPoint goodArray[getNumberOfKeypoints()];
assert(validateKeypoints(kpts, img->width, img->height));
// Allocate space for the integral image:
allocateIntegralImage(img);
// Calculate the integral image:
cvIntegral(img, integralImage, NULL, NULL);
printf("NumberOfKeypoints : %d\n", getNumberOfKeypoints());
// Iterate over keypoints:
for (i = 0; i < getNumberOfKeypoints(); i++) {
int inWS = integralImage->step / CV_ELEM_SIZE(integralImage->type);
int* iD = integralImage->data.i;
int j;
for (j = 0; j < DESC_LEN; ++j) {
const pair tL = testLocations[j];
const CvPoint p1 = CV_POINT_PLUS(kpts[i], tL.point1);
const CvPoint p2 = CV_POINT_PLUS(kpts[i], tL.point2);
const int intensity1 = GET_PIXEL_NW(iD, p1, inWS)
- GET_PIXEL_NE(iD, p1, inWS) - GET_PIXEL_SW(iD, p1, inWS)
+ GET_PIXEL_SE(iD, p1, inWS);
const int intensity2 = GET_PIXEL_NW(iD, p2, inWS)
- GET_PIXEL_NE(iD, p2, inWS) - GET_PIXEL_SW(iD, p2, inWS)
+ GET_PIXEL_SE(iD, p2, inWS);
if (intensity1 < intensity2) {
descriptors[i][j] = 1;
} else {
descriptors[i][j] = 0;
}
}
}
}
//Parallel version
void
OMP_eliminateJunction(IplImage* iplEdge, uint8 edgC, uint8 backC)
{
int W = iplEdge->width;
int H = iplEdge->height;
Image<uint8> iplEdgeT(iplEdge);
IplImage* iplIntImg = cvCreateImage( cvSize(W+1, H+1), IPL_DEPTH_32S, 1 );
cvIntegral(iplEdge, iplIntImg,NULL, NULL );
Image<int> iplIntImgT(iplIntImg);
int nthreads = OMP_getNbThread();
int tid;
#pragma omp parallel private(tid)
{
tid = omp_get_thread_num();
int bg = tid+1;
printf("[%d] Starting at %d\n", tid,bg);
for (int y = bg; y < iplEdge->height-1; y+=nthreads) {
//printf("%d %d %d %d\n", tid,y,iplEdge->height-1, nthreads );
for (int x = 1; x < iplEdge->width-1; x++) {
//even if current pixel is not mark, we mark it
//a junction can have no edge!
//double s = iplEdgeT[y][x];
//if (s!=edgC)
// nbEdge+=255;
//faster than above
uint nbEdge=255 - iplEdgeT[y][x]; //if no edge, add 255, otherwise, start at 0, will be added below
nbEdge += ( iplIntImgT[y-1][x-1]
+ iplIntImgT[y+2][x+2]
- iplIntImgT[y+2][x-1]
- iplIntImgT[y-1][x+2]);
if (nbEdge>=4*255)
iplEdgeT[y][x] = backC;
}}
} //end pragma omp parallel private(tid)
}
/**
* Returns Haar feature values
* Precondition: image should be 64x64
*/
void HaarFeatures::getFeatureValues(vector<double>& feature_values,
const IplImage* img) const {
//Image should be 64x64 (because the Haar features are based on 64x64 images)
assert(img->width == 64);
assert(img->height == 64);
IplImage *iImage = cvCreateImage(cvSize(img->width + 1, img->height + 1), IPL_DEPTH_32S, 1);
cvIntegral(img, iImage);
for(vector<HaarFeature>::size_type i = 0; i < features.size(); i++) {
double feature_value = computeFeatureValueIntegral(iImage, features[i]);
feature_values.push_back(feature_value);
//printFeature(features[i]);
//cerr << "Feature value [" << i << "]: " << feature_value << endl;
}
cvReleaseImage(&iImage);
}
CLIFIntegralResult
clifIntegral(const IplImage* source,
CLIFEnvironmentData* data,
const cl_bool use_opencl)
{
CLIFIntegralResult ret;
if(!use_opencl) {
ret.image = cvCreateMat(source->height + 1, source->width + 1, CV_32SC1);
ret.square_image = cvCreateMat(source->height + 1, source->width + 1, CV_64FC1);
cvIntegral(source, ret.image, ret.square_image);
return ret;
}
cl_int error = CL_SUCCESS;
// Init buffer
error = clEnqueueWriteBuffer(data->environment.queue, data->integral_image_data.buffers[0], CL_FALSE, 0, source->width * source->height, source, 0, NULL, NULL);
clCheckOrExit(error);
// Run sum rows kernel
error = clEnqueueNDRangeKernel(data->environment.queue, data->environment.kernels[1], 1, NULL, &(data->integral_image_data.global_size[0]), &(data->integral_image_data.local_size[0]), 0, NULL, NULL);
clCheckOrExit(error);
// Run sum cols kernel
error = clEnqueueNDRangeKernel(data->environment.queue, data->environment.kernels[2], 1, NULL, &(data->integral_image_data.global_size[1]), &(data->integral_image_data.local_size[1]), 0, NULL, NULL);
clCheckOrExit(error);
// Read result
cl_uint* result = (cl_uint*)clEnqueueMapBuffer(data->environment.queue, data->integral_image_data.buffers[3], CL_TRUE, CL_MAP_READ, 0, (source->width + 1) * (source->height + 1) * sizeof(cl_uint), 0, NULL, NULL, &error);
clCheckOrExit(error);
cl_ulong* square_result = (cl_ulong*)clEnqueueMapBuffer(data->environment.queue, data->integral_image_data.buffers[4], CL_TRUE, CL_MAP_READ, 0, (source->width + 1) * (source->height + 1) * sizeof(cl_ulong), 0, NULL, NULL, &error);
clCheckOrExit(error);
data->integral_image_data.ptr = result;
// Return
ret.image = cvCreateMat(source->height + 1, source->width + 1, CV_32SC1);
cvSetData(ret.image, result, (source->width + 1) * sizeof(cl_uint));
ret.square_image = cvCreateMatHeader(source->height + 1, source->width + 1, CV_64FC1);
cvSetData(ret.square_image, square_result, (source->width + 1) * sizeof(cl_ulong));
return ret;
}
CV_IMPL void
cvMatchTemplate( const CvArr* _img, const CvArr* _templ, CvArr* _result, int method )
{
cv::Ptr<CvMat> sum, sqsum;
int coi1 = 0, coi2 = 0;
int depth, cn;
int i, j, k;
CvMat stub, *img = (CvMat*)_img;
CvMat tstub, *templ = (CvMat*)_templ;
CvMat rstub, *result = (CvMat*)_result;
CvScalar templ_mean = cvScalarAll(0);
double templ_norm = 0, templ_sum2 = 0;
int idx = 0, idx2 = 0;
double *p0, *p1, *p2, *p3;
double *q0, *q1, *q2, *q3;
double inv_area;
int sum_step, sqsum_step;
int num_type = method == CV_TM_CCORR || method == CV_TM_CCORR_NORMED ? 0 :
method == CV_TM_CCOEFF || method == CV_TM_CCOEFF_NORMED ? 1 : 2;
int is_normed = method == CV_TM_CCORR_NORMED ||
method == CV_TM_SQDIFF_NORMED ||
method == CV_TM_CCOEFF_NORMED;
img = cvGetMat( img, &stub, &coi1 );
templ = cvGetMat( templ, &tstub, &coi2 );
result = cvGetMat( result, &rstub );
if( CV_MAT_DEPTH( img->type ) != CV_8U &&
CV_MAT_DEPTH( img->type ) != CV_32F )
CV_Error( CV_StsUnsupportedFormat,
"The function supports only 8u and 32f data types" );
if( !CV_ARE_TYPES_EQ( img, templ ))
CV_Error( CV_StsUnmatchedSizes, "image and template should have the same type" );
if( CV_MAT_TYPE( result->type ) != CV_32FC1 )
CV_Error( CV_StsUnsupportedFormat, "output image should have 32f type" );
if( img->rows < templ->rows || img->cols < templ->cols )
{
CvMat* t;
CV_SWAP( img, templ, t );
}
if( result->rows != img->rows - templ->rows + 1 ||
result->cols != img->cols - templ->cols + 1 )
CV_Error( CV_StsUnmatchedSizes, "output image should be (W - w + 1)x(H - h + 1)" );
if( method < CV_TM_SQDIFF || method > CV_TM_CCOEFF_NORMED )
CV_Error( CV_StsBadArg, "unknown comparison method" );
depth = CV_MAT_DEPTH(img->type);
cn = CV_MAT_CN(img->type);
icvCrossCorr( img, templ, result );
if( method == CV_TM_CCORR )
return;
inv_area = 1./((double)templ->rows * templ->cols);
sum = cvCreateMat( img->rows + 1, img->cols + 1, CV_MAKETYPE( CV_64F, cn ));
if( method == CV_TM_CCOEFF )
{
cvIntegral( img, sum, 0, 0 );
templ_mean = cvAvg( templ );
q0 = q1 = q2 = q3 = 0;
}
else
{
CvScalar _templ_sdv = cvScalarAll(0);
sqsum = cvCreateMat( img->rows + 1, img->cols + 1, CV_MAKETYPE( CV_64F, cn ));
cvIntegral( img, sum, sqsum, 0 );
cvAvgSdv( templ, &templ_mean, &_templ_sdv );
templ_norm = CV_SQR(_templ_sdv.val[0]) + CV_SQR(_templ_sdv.val[1]) +
CV_SQR(_templ_sdv.val[2]) + CV_SQR(_templ_sdv.val[3]);
if( templ_norm < DBL_EPSILON && method == CV_TM_CCOEFF_NORMED )
{
cvSet( result, cvScalarAll(1.) );
return;
}
templ_sum2 = templ_norm +
CV_SQR(templ_mean.val[0]) + CV_SQR(templ_mean.val[1]) +
CV_SQR(templ_mean.val[2]) + CV_SQR(templ_mean.val[3]);
if( num_type != 1 )
{
templ_mean = cvScalarAll(0);
templ_norm = templ_sum2;
}
templ_sum2 /= inv_area;
templ_norm = sqrt(templ_norm);
templ_norm /= sqrt(inv_area); // care of accuracy here
q0 = (double*)sqsum->data.ptr;
q1 = q0 + templ->cols*cn;
q2 = (double*)(sqsum->data.ptr + templ->rows*sqsum->step);
q3 = q2 + templ->cols*cn;
}
p0 = (double*)sum->data.ptr;
p1 = p0 + templ->cols*cn;
p2 = (double*)(sum->data.ptr + templ->rows*sum->step);
p3 = p2 + templ->cols*cn;
sum_step = sum ? sum->step / sizeof(double) : 0;
sqsum_step = sqsum ? sqsum->step / sizeof(double) : 0;
for( i = 0; i < result->rows; i++ )
{
float* rrow = (float*)(result->data.ptr + i*result->step);
idx = i * sum_step;
idx2 = i * sqsum_step;
for( j = 0; j < result->cols; j++, idx += cn, idx2 += cn )
{
double num = rrow[j], t;
double wnd_mean2 = 0, wnd_sum2 = 0;
if( num_type == 1 )
{
for( k = 0; k < cn; k++ )
{
t = p0[idx+k] - p1[idx+k] - p2[idx+k] + p3[idx+k];
wnd_mean2 += CV_SQR(t);
num -= t*templ_mean.val[k];
}
wnd_mean2 *= inv_area;
}
if( is_normed || num_type == 2 )
{
for( k = 0; k < cn; k++ )
{
t = q0[idx2+k] - q1[idx2+k] - q2[idx2+k] + q3[idx2+k];
wnd_sum2 += t;
}
if( num_type == 2 )
num = wnd_sum2 - 2*num + templ_sum2;
}
if( is_normed )
{
t = sqrt(MAX(wnd_sum2 - wnd_mean2,0))*templ_norm;
if( fabs(num) < t )
num /= t;
else if( fabs(num) < t*1.125 )
num = num > 0 ? 1 : -1;
else
num = method != CV_TM_SQDIFF_NORMED ? 0 : 1;
}
rrow[j] = (float)num;
}
}
}
CV_IMPL void
cvMatchTemplate( const CvArr* _img, const CvArr* _templ, CvArr* _result, int method )
{
CvMat* sum = 0;
CvMat* sqsum = 0;
CV_FUNCNAME( "cvMatchTemplate" );
__BEGIN__;
int coi1 = 0, coi2 = 0;
int depth, cn;
int i, j, k;
CvMat stub, *img = (CvMat*)_img;
CvMat tstub, *templ = (CvMat*)_templ;
CvMat rstub, *result = (CvMat*)_result;
CvScalar templ_mean = cvScalarAll(0);
double templ_norm = 0, templ_sum2 = 0;
int idx = 0, idx2 = 0;
double *p0, *p1, *p2, *p3;
double *q0, *q1, *q2, *q3;
double inv_area;
int sum_step, sqsum_step;
int num_type = method == CV_TM_CCORR || method == CV_TM_CCORR_NORMED ? 0 :
method == CV_TM_CCOEFF || method == CV_TM_CCOEFF_NORMED ? 1 : 2;
int is_normed = method == CV_TM_CCORR_NORMED ||
method == CV_TM_SQDIFF_NORMED ||
method == CV_TM_CCOEFF_NORMED;
CV_CALL( img = cvGetMat( img, &stub, &coi1 ));
CV_CALL( templ = cvGetMat( templ, &tstub, &coi2 ));
CV_CALL( result = cvGetMat( result, &rstub ));
if( CV_MAT_DEPTH( img->type ) != CV_8U &&
CV_MAT_DEPTH( img->type ) != CV_32F )
CV_ERROR( CV_StsUnsupportedFormat,
"The function supports only 8u and 32f data types" );
if( !CV_ARE_TYPES_EQ( img, templ ))
CV_ERROR( CV_StsUnmatchedSizes, "image and template should have the same type" );
if( CV_MAT_TYPE( result->type ) != CV_32FC1 )
CV_ERROR( CV_StsUnsupportedFormat, "output image should have 32f type" );
if( img->rows < templ->rows || img->cols < templ->cols )
{
CvMat* t;
CV_SWAP( img, templ, t );
}
if( result->rows != img->rows - templ->rows + 1 ||
result->cols != img->cols - templ->cols + 1 )
CV_ERROR( CV_StsUnmatchedSizes, "output image should be (W - w + 1)x(H - h + 1)" );
if( method < CV_TM_SQDIFF || method > CV_TM_CCOEFF_NORMED )
CV_ERROR( CV_StsBadArg, "unknown comparison method" );
depth = CV_MAT_DEPTH(img->type);
cn = CV_MAT_CN(img->type);
if( is_normed && cn == 1 && templ->rows > 8 && templ->cols > 8 &&
img->rows > templ->cols && img->cols > templ->cols )
{
CvTemplMatchIPPFunc ipp_func =
depth == CV_8U ?
(method == CV_TM_SQDIFF_NORMED ? (CvTemplMatchIPPFunc)icvSqrDistanceValid_Norm_8u32f_C1R_p :
method == CV_TM_CCORR_NORMED ? (CvTemplMatchIPPFunc)icvCrossCorrValid_Norm_8u32f_C1R_p :
(CvTemplMatchIPPFunc)icvCrossCorrValid_NormLevel_8u32f_C1R_p) :
(method == CV_TM_SQDIFF_NORMED ? (CvTemplMatchIPPFunc)icvSqrDistanceValid_Norm_32f_C1R_p :
method == CV_TM_CCORR_NORMED ? (CvTemplMatchIPPFunc)icvCrossCorrValid_Norm_32f_C1R_p :
(CvTemplMatchIPPFunc)icvCrossCorrValid_NormLevel_32f_C1R_p);
if( ipp_func )
{
CvSize img_size = cvGetMatSize(img), templ_size = cvGetMatSize(templ);
IPPI_CALL( ipp_func( img->data.ptr, img->step ? img->step : CV_STUB_STEP,
img_size, templ->data.ptr,
templ->step ? templ->step : CV_STUB_STEP,
templ_size, result->data.ptr,
result->step ? result->step : CV_STUB_STEP ));
for( i = 0; i < result->rows; i++ )
{
float* rrow = (float*)(result->data.ptr + i*result->step);
for( j = 0; j < result->cols; j++ )
{
if( fabs(rrow[j]) > 1. )
rrow[j] = rrow[j] < 0 ? -1.f : 1.f;
}
}
EXIT;
}
}
CV_CALL( icvCrossCorr( img, templ, result ));
if( method == CV_TM_CCORR )
EXIT;
inv_area = 1./((double)templ->rows * templ->cols);
CV_CALL( sum = cvCreateMat( img->rows + 1, img->cols + 1,
CV_MAKETYPE( CV_64F, cn )));
if( method == CV_TM_CCOEFF )
{
CV_CALL( cvIntegral( img, sum, 0, 0 ));
CV_CALL( templ_mean = cvAvg( templ ));
q0 = q1 = q2 = q3 = 0;
}
else
{
CvScalar _templ_sdv = cvScalarAll(0);
CV_CALL( sqsum = cvCreateMat( img->rows + 1, img->cols + 1,
CV_MAKETYPE( CV_64F, cn )));
CV_CALL( cvIntegral( img, sum, sqsum, 0 ));
CV_CALL( cvAvgSdv( templ, &templ_mean, &_templ_sdv ));
templ_norm = CV_SQR(_templ_sdv.val[0]) + CV_SQR(_templ_sdv.val[1]) +
CV_SQR(_templ_sdv.val[2]) + CV_SQR(_templ_sdv.val[3]);
if( templ_norm < DBL_EPSILON && method == CV_TM_CCOEFF_NORMED )
{
cvSet( result, cvScalarAll(1.) );
EXIT;
}
templ_sum2 = templ_norm +
CV_SQR(templ_mean.val[0]) + CV_SQR(templ_mean.val[1]) +
CV_SQR(templ_mean.val[2]) + CV_SQR(templ_mean.val[3]);
if( num_type != 1 )
{
templ_mean = cvScalarAll(0);
templ_norm = templ_sum2;
}
templ_sum2 /= inv_area;
templ_norm = sqrt(templ_norm);
templ_norm /= sqrt(inv_area); // care of accuracy here
q0 = (double*)sqsum->data.ptr;
q1 = q0 + templ->cols*cn;
q2 = (double*)(sqsum->data.ptr + templ->rows*sqsum->step);
q3 = q2 + templ->cols*cn;
}
p0 = (double*)sum->data.ptr;
p1 = p0 + templ->cols*cn;
p2 = (double*)(sum->data.ptr + templ->rows*sum->step);
p3 = p2 + templ->cols*cn;
sum_step = sum ? sum->step / sizeof(double) : 0;
sqsum_step = sqsum ? sqsum->step / sizeof(double) : 0;
for( i = 0; i < result->rows; i++ )
{
float* rrow = (float*)(result->data.ptr + i*result->step);
idx = i * sum_step;
idx2 = i * sqsum_step;
for( j = 0; j < result->cols; j++, idx += cn, idx2 += cn )
{
double num = rrow[j], t;
double wnd_mean2 = 0, wnd_sum2 = 0;
if( num_type == 1 )
{
for( k = 0; k < cn; k++ )
{
t = p0[idx+k] - p1[idx+k] - p2[idx+k] + p3[idx+k];
wnd_mean2 += CV_SQR(t);
num -= t*templ_mean.val[k];
}
wnd_mean2 *= inv_area;
}
if( is_normed || num_type == 2 )
{
for( k = 0; k < cn; k++ )
{
t = q0[idx2+k] - q1[idx2+k] - q2[idx2+k] + q3[idx2+k];
wnd_sum2 += t;
}
if( num_type == 2 )
num = wnd_sum2 - 2*num + templ_sum2;
}
if( is_normed )
{
t = sqrt(MAX(wnd_sum2 - wnd_mean2,0))*templ_norm;
if( t > DBL_EPSILON )
{
num /= t;
if( fabs(num) > 1. )
num = num > 0 ? 1 : -1;
}
else
num = method != CV_TM_SQDIFF_NORMED || num < DBL_EPSILON ? 0 : 1;
}
rrow[j] = (float)num;
}
}
__END__;
cvReleaseMat( &sum );
cvReleaseMat( &sqsum );
}
*/
int main(int argc, char *argv[])
{
DSP_cvStartDSP();
CvCapture * capture;
IplImage *videoFrame, *convFrame, *convOpencvFrame, *unsignedFrame;
IplImage *dataImage, *integralImage;
int key;
int *ptr;
float *flPtr;
int i,j;
float tempFloat=0.0;
float *floatDataPtr;
float *floatOutPtr;
/* Data to test cvIntegral() */
unsigned char intdata[] = { 151, 57, 116, 170, 9, 247, 208, 140, 150, 60, 88, 77, 4, 6, 162, 6,
31, 143, 178, 3, 135, 91, 54, 154, 193, 161, 20, 162, 137, 150, 128, 224,
214, 113, 9, 28, 53, 211, 98, 217, 149, 233, 231, 127, 115, 203, 177, 42,
62, 155, 3, 103, 127, 16, 135, 131, 211, 158, 9, 2, 106, 227, 249, 255 }; //16 x 4
if ( argc < 2 ) {
printf( "Usage: ./remote_ti_platforms_evm3530_opencv.xv5T [option] \n");
printf("option:\ni. integral\ns. sobel\nd. dft\n");
printf("Following are the all usage:\n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T i dsp(To test integral-image algorithm using DSP. Input is from webcam). Note: You need to install VLIB to test this.\n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T i arm(To test integral-image algorithm using ARM. Input is from webcam). \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T i test(To test integral-image algorithm using test data given in APP. Input is from webcam). \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T d dsp (To test DFT algorithm using DSP) \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T d arm (To test DFT algorithm using ARM) \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T s tree.avi dsp (To test sobel algorithm for movie clip tree.avi using DSP) \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T s tree.avi arm (To test sobel algorithm for movie clip tree.avi using ARM) \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T s webcam dsp (To test sobel algorithm for image from webcam using DSP) \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T s webcam arm (To test sobel algorithm for image from webcam using ARM) \n");
printf("./remote_ti_platforms_evm3530_opencv.xv5T rgb2gray (To test RGB to Gray for image from webcam.) \n");
return (-1);
}
if (*argv[1] == 's' && argc < 3) {
printf( "Usage: ./remote_ti_platforms_evm3530_opencv.xv5T s tree.avi \n");
printf( "Usage: ./remote_ti_platforms_evm3530_opencv.xv5T s webcam \n");
return (-1);
}
switch (*argv[1]) {
case 'i':
switch (*argv[2]) {
case 'd': { //'d' dor DSP accelerated
cvNamedWindow( "video", CV_WINDOW_AUTOSIZE );
capture = cvCaptureFromCAM(-1);
if ( !capture) {
printf("Error: Video capture initialization failed.\n");
break;
}
videoFrame = cvQueryFrame ( capture );
if ( !videoFrame) {
printf("**Error reading from webcam\n");
break;
}
/* create new image for the grayscale version */
convFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* create sobel filtered image */
convOpencvFrame = cvCreateImage( cvSize( convFrame->width+1, convFrame->height+1 ), IPL_DEPTH_32S, 1 );
/* Process the first frame outside the loop*/
DSP_cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
DSP_cvSyncDSP();
while ( key != 'q') {
/* Time to test and benchmark DSP based sobel filter */
Time_reset(&sTime);
Time_delta(&sTime,&time);
/*Find integral image */
DSP_cvIntegral(convFrame,convOpencvFrame,NULL,NULL);
/* get next frame */
videoFrame = cvQueryFrame( capture);
if ( !videoFrame) {
printf("***The End***\n");
break;
}
DSP_cvSyncDSP();
/* Do color conversion */
DSP_cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
/* show Image */ //Since I am using VLIB for IntegralImage, its output image width and height is same as source image.
convOpencvFrame->width -= 1; convOpencvFrame->height -= 1;
convOpencvFrame->widthStep -= sizeof(int) * convOpencvFrame->nChannels;
cvShowImage("video", convOpencvFrame);
convOpencvFrame->width += 1; convOpencvFrame->height += 1;
convOpencvFrame->widthStep += sizeof(int) * convOpencvFrame->nChannels;
/* Sync with DSP */
DSP_cvSyncDSP();
Time_delta(&sTime,&time);
printf("Total execution time = %dus\n",(unsigned int)time);
key = cvWaitKey( 15 );
}
cvDestroyWindow("video");
cvReleaseImage(&videoFrame);
cvReleaseImage(&convFrame);
cvReleaseImage(&convOpencvFrame);
cvReleaseCapture(&capture);
}
break; /* End of sobel test */
case 'a': { // 'a' for ARM side
cvNamedWindow( "video", CV_WINDOW_AUTOSIZE );
capture = cvCaptureFromCAM(-1);
if ( !capture) {
printf("Error: Video capture initialization failed.\n");
break;
}
videoFrame = cvQueryFrame ( capture );
if ( !videoFrame) {
printf("**Error reading from webcam\n");
break;
}
/* create new image for the grayscale version */
convFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* create sobel filtered image */
convOpencvFrame = cvCreateImage( cvSize( convFrame->width+1, convFrame->height+1 ), IPL_DEPTH_32S, 1 );
/* Process the first frame outside the loop*/
cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
while ( key != 'q') {
/* Time to test and benchmark DSP based sobel filter */
Time_reset(&sTime);
Time_delta(&sTime,&time);
/*Find integral image */
cvIntegral(convFrame,convOpencvFrame,NULL,NULL);
/* get next frame */
videoFrame = cvQueryFrame( capture);
if ( !videoFrame) {
printf("***The End***\n");
break;
}
/* Do color conversion */
cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
/* show Image */
cvShowImage("video", convOpencvFrame);
Time_delta(&sTime,&time);
printf("Total execution time = %dus\n",(unsigned int)time);
key = cvWaitKey( 15 );
}
cvDestroyWindow("video");
cvReleaseImage(&videoFrame);
cvReleaseImage(&convFrame);
cvReleaseImage(&convOpencvFrame);
cvReleaseCapture(&capture);
}
break; /* End of sobel test */
case 't': { /* This to test the consistency of algorithm */
/* Start of integral image test */
dataImage = cvCreateImageHeader(cvSize(16, 4), IPL_DEPTH_8U, 1);
integralImage = cvCreateImageHeader(cvSize(17, 5), IPL_DEPTH_32S, 1);
unsigned char *data = (unsigned char*)cvAlloc(16*4*sizeof(char));
unsigned int *sum = (unsigned int *)cvAlloc(17*5*sizeof(int));
memcpy(data, &intdata[0], 16*4*sizeof(char));
memset(sum, 0, 17*5*sizeof(int));
dataImage->imageData = ( char *)data;
integralImage->imageData = ( char *)sum;
/* DSP based integral */
DSP_cvIntegral(dataImage,integralImage,NULL,NULL);
ptr = (int *)integralImage->imageData;
printf(" The integral image is:\n");
for (i=0;i<4;i++){
for (j=0;j<16;j++){
printf("%d \t ", ptr[i* 16 + j]);
}
printf("\n");
}
/* Arm based cvIntegral() */
cvIntegral(dataImage, integralImage,NULL,NULL);
ptr = (int *)integralImage->imageData;
printf(" The integral image is:\n");
for (i=0;i<5;i++){
for (j=0;j<17;j++){
printf("%d \t ", ptr[i* 17 + j]);
}
printf("\n");
}
cvFree(&dataImage);
cvFree(&integralImage);
}
break; /* End of integral image test */
default:
printf("Input argument Error:\n Usage :\n./remote_ti_platforms_evm3530_opencv.xv5T i d \n./remote_ti_platforms_evm3530_opencv.xv5T i a\n ./remote_ti_platforms_evm3530_opencv.xv5T i d \n");
break;
}
break;
/* Start RGB to Y test */
case 'r': {
cvNamedWindow( "video", CV_WINDOW_AUTOSIZE );
capture = cvCaptureFromCAM(-1);
if ( !capture) {
printf("Error: Video capture initialization failed.\n");
break;
}
videoFrame = cvQueryFrame ( capture );
if ( !videoFrame) {
printf("**Error reading from webcam\n");
break;
}
/* create new image for the grayscale version */
convFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
unsignedFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
while ( key != 'q') {
/* Time to test and benchmark DSP based sobel filter */
Time_reset(&sTime);
Time_delta(&sTime,&time);
/* Process the first frame outside the loop*/
DSP_cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
videoFrame = cvQueryFrame( capture);
if ( !videoFrame) {
printf("***The End***\n");
break;
}
DSP_cvSyncDSP();
cvShowImage("video",convFrame);
Time_delta(&sTime,&time);
printf("Total execution time1 = %dus\n",(unsigned int)time);
key = cvWaitKey( 15 );
}
cvDestroyWindow("video");
cvReleaseImage(&videoFrame);
cvReleaseImage(&convFrame);
}
break;
case 's': /* Start of sobel test */
switch (*argv[2]) {
case 't':
switch(*argv[3]) {
case 'd': { //'d' dor DSP accelerated
cvNamedWindow( "video", CV_WINDOW_AUTOSIZE );
capture = cvCreateFileCapture(argv[2]);
if ( !capture) {
printf("Error: Video not found\n");
break;
}
videoFrame = cvQueryFrame ( capture );
if ( !videoFrame) {
printf("**Error reading video\n");
break;
}
/* create new image for the grayscale version */
convFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* create sobel filtered image */
convOpencvFrame = cvCreateImage( cvSize( convFrame->width, convFrame->height ), IPL_DEPTH_16S, 1 );
DSP_cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
DSP_cvSyncDSP();
while ( key != 'q') {
/* Time to test and benchmark DSP based sobel filter */
Time_reset(&sTime);
Time_delta(&sTime,&time);
DSP_cvSobel(convFrame,convOpencvFrame,1,1,3);
/* get next frame */
videoFrame = cvQueryFrame( capture);
if ( !videoFrame) {
printf("***The End***\n");
break;
}
/* Sync with DSP*/
DSP_cvSyncDSP();
/* Start RGB To Y conversion of next frame */
DSP_cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
/* Display Filtered Image */
cvShowImage("video", convOpencvFrame);
/* Sync DSP */
DSP_cvSyncDSP();
Time_delta(&sTime,&time);
printf("Total execution time = %dus\n",(unsigned int)time);
key = cvWaitKey( 15 );
}
cvDestroyWindow("video");
cvReleaseImage(&videoFrame);
cvReleaseImage(&convFrame);
cvReleaseImage(&convOpencvFrame);
}
break; /* End of sobel test */
case 'a': { // 'a' for ARM side
cvNamedWindow( "video", CV_WINDOW_AUTOSIZE );
capture = cvCreateFileCapture(argv[2]);
if ( !capture) {
printf("Error: Video not found\n");
break;
}
videoFrame = cvQueryFrame ( capture );
if ( !videoFrame) {
printf("**Error reading video\n");
break;
}
/* create new image for the grayscale version */
convFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* create sobel filtered image */
convOpencvFrame = cvCreateImage( cvSize( convFrame->width, convFrame->height ), IPL_DEPTH_16S, 1 );
cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
while ( key != 'q') {
/* Time to test and benchmark DSP based sobel filter */
Time_reset(&sTime);
Time_delta(&sTime,&time);
cvSobel(convFrame,convOpencvFrame,1,1,3);
/* get next frame */
videoFrame = cvQueryFrame( capture);
if ( !videoFrame) {
printf("***The End***\n");
break;
}
/* Start RGB To Y conversion of next frame */
cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
/* Display Filtered Image */
cvShowImage("video", convOpencvFrame);
Time_delta(&sTime,&time);
printf("Total execution time = %dus\n",(unsigned int)time);
key = cvWaitKey( 15 );
}
cvDestroyWindow("video");
cvReleaseImage(&videoFrame);
cvReleaseImage(&convFrame);
cvReleaseImage(&convOpencvFrame);
}
break; /* End of sobel test */
default:
printf("Input argument Error:\n Usage :\n./remote_ti_platforms_evm3530_opencv.xv5T s tree.avi d \n./remote_ti_platforms_evm3530_opencv.xv5T s tree.avi a\n");
}
break;
case 'w':
switch(*argv[3]) {
case 'd': { //'d' dor DSP accelerated
cvNamedWindow( "video", CV_WINDOW_AUTOSIZE );
capture = cvCaptureFromCAM(-1);
if ( !capture) {
printf("Error: Video capture initialization failed.\n");
break;
}
videoFrame = cvQueryFrame ( capture );
if ( !videoFrame) {
printf("**Error reading from webcam\n");
break;
}
/* create new image for the grayscale version */
convFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* create sobel filtered image */
convOpencvFrame = cvCreateImage( cvSize( convFrame->width, convFrame->height ), IPL_DEPTH_16S, 1 );
/* Create unsigned image */
unsignedFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* Process the first frame outside the loop*/
DSP_cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
DSP_cvSyncDSP();
while ( key != 'q') {
/* Time to test and benchmark DSP based sobel filter */
Time_reset(&sTime);
Time_delta(&sTime,&time);
DSP_cvSobel(convFrame,convOpencvFrame,1,1,3);
/* get next frame */
videoFrame = cvQueryFrame( capture);
if ( !videoFrame) {
printf("***The End***\n");
break;
}
/* Sync with DSP*/
DSP_cvSyncDSP();
/* Start RGB To Y conversion of next frame */
DSP_cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
/* Convert signed image to unsign image for better clarity */
cvConvert(convOpencvFrame,unsignedFrame);
/* Display Filtered Image */
cvShowImage("video", unsignedFrame);
/* Sync DSP */
DSP_cvSyncDSP();
Time_delta(&sTime,&time);
printf("Total execution time = %dus\n",(unsigned int)time);
key = cvWaitKey( 15 );
}
cvDestroyWindow("video");
cvReleaseImage(&videoFrame);
cvReleaseImage(&convFrame);
cvReleaseImage(&convOpencvFrame);
}
break; /* End of sobel test */
case 'a': { // 'a' for ARM side
cvNamedWindow( "video", CV_WINDOW_AUTOSIZE );
capture = cvCaptureFromCAM(-1);
if ( !capture) {
printf("Error: Video capture initialization failed.\n");
break;
}
videoFrame = cvQueryFrame ( capture );
if ( !videoFrame) {
printf("**Error reading from webcam\n");
break;
}
/* create new image for the grayscale version */
convFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* create sobel filtered image */
convOpencvFrame = cvCreateImage( cvSize( convFrame->width, convFrame->height ), IPL_DEPTH_16S, 1 );
/* Create unsigned image */
unsignedFrame = cvCreateImage( cvSize( videoFrame->width, videoFrame->height ), IPL_DEPTH_8U, 1 );
/* Process the first frame outside the loop*/
cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
while ( key != 'q') {
/* Time to test and benchmark DSP based sobel filter */
Time_reset(&sTime);
Time_delta(&sTime,&time);
cvSobel(convFrame,convOpencvFrame,1,1,3);
/* get next frame */
videoFrame = cvQueryFrame( capture);
if ( !videoFrame) {
printf("***The End***\n");
break;
}
/* Start RGB To Y conversion of next frame */
cvCvtColor(videoFrame,convFrame,CV_RGB2GRAY);
/* Convert signed image to unsign image for better clarity */
cvConvert(convOpencvFrame,unsignedFrame);
/* Display Filtered Image */
cvShowImage("video", unsignedFrame);
Time_delta(&sTime,&time);
printf("Total execution time = %dus\n",(unsigned int)time);
key = cvWaitKey( 15 );
}
cvDestroyWindow("video");
cvReleaseImage(&videoFrame);
cvReleaseImage(&convFrame);
cvReleaseImage(&convOpencvFrame);
}
break; /* End of sobel test */
default:
printf("Input argument Error:\n Usage :\n./remote_ti_platforms_evm3530_opencv.xv5T s webcam d \n./remote_ti_platforms_evm3530_opencv.xv5T s webcam a\n");
}
break;
default:
printf("Input argument Error:\n Usage :\n./remote_ti_platforms_evm3530_opencv.xv5T s tree.avi \n./remote_ti_platforms_evm3530_opencv.xv5T s webcam\n");
break;
}
break;
case 'd':
switch(*argv[2]) {
case 'd': { //'d' dor DSP accelerated
int row =63;
int col =63;
floatDataPtr = (float *)cvAlloc(sizeof(float)*row*col* 2);
floatOutPtr = (float *)cvAlloc(sizeof(float)*row*col * 2);
dataImage = cvCreateImageHeader(cvSize(col, row), IPL_DEPTH_32F, 2);
dataImage->imageData = (char *)floatDataPtr;
integralImage = cvCreateImageHeader(cvSize(col, row), IPL_DEPTH_32F, 2);
integralImage->imageData = (char *)floatOutPtr;
for (i=0; i< row * col * 2; i+=2) {
tempFloat += 1.0;
floatDataPtr[i] = tempFloat;
floatDataPtr[i+1] = 0.0;
}
Time_reset(&sTime);
Time_delta(&sTime,&time);
DSP_cvDFT(dataImage,integralImage,CV_DXT_FORWARD,0);
Time_delta(&sTime,&time);
DSP_cvSyncDSP();
/* Print the output data for DFT */
flPtr = (float *)integralImage->imageData;
printf(" The DFT output is:\n");
for (i=0;i<integralImage->height;i++){
key = 0;
for (j=0;j<integralImage->width * 2;j+=2){
key++;
printf("%f + i%f\t", flPtr[(i*integralImage->width * 2)+j], flPtr[(i*integralImage->width * 2) + j + 1]);
if ((key % 5) == 0) printf("\n");
}
printf("\n");
}
printf("DSP_cvDFT Total execution time = %dus\n",(unsigned int)time);
cvFree(&floatDataPtr);
cvFree(&floatOutPtr);
}
break;
case 'a': {
int row =63;
int col =63;
floatDataPtr = (float *)cvAlloc(sizeof(float)*row*col*2);
floatOutPtr = (float *)cvAlloc(sizeof(float)*row*col*2);
dataImage = cvCreateImageHeader(cvSize(col, row), IPL_DEPTH_32F, 2);
dataImage->imageData = (char *)floatDataPtr;
integralImage = cvCreateImageHeader(cvSize(col, row), IPL_DEPTH_32F, 2);
integralImage->imageData = (char *)floatOutPtr;
for (i=0; i< row * col * 2; i+=2) {
tempFloat += 1.0;
floatDataPtr[i] = tempFloat;
floatDataPtr[i+1] = 0.0;
}
Time_reset(&sTime);
Time_delta(&sTime,&time);
cvDFT(dataImage,integralImage,CV_DXT_FORWARD,0);
Time_delta(&sTime,&time);
/* Print the output data for DFT */
flPtr = (float *)integralImage->imageData;
printf(" The DFT output is:\n");
for (i=0;i<integralImage->height;i++){
key = 0;
for (j=0;j<integralImage->width * 2;j+=2){
key++;
printf("%f + i%f\t", flPtr[(i*integralImage->width * 2)+j], flPtr[(i*integralImage->width * 2) + j + 1]);
if ((key % 5) == 0) printf("\n");
}
printf("\n");
}
printf("cvDFT Total execution time = %dus\n",(unsigned int)time);
cvFree(&floatDataPtr);
cvFree(&floatOutPtr);
}
break;
default:
printf("Input argument Error:\n Usage :\n./remote_ti_platforms_evm3530_opencv.xv5T d d \n./remote_ti_platforms_evm3530_opencv.xv5T d a\n");
} // end of latest switch(*argv[3])
break;
default:
return -1;
}
DSP_cvEndDSP();
return 0;
int main( int argc, char** argv )
{
int wcam = -1;
CvCapture* capture = 0;
IplImage *frame = 0;
IplImage *frame_resized = 0;
IplImage *frame_resized2 = 0;
CvSize min_window_size, max_window_size;
min_window_size.width = 40;
min_window_size.height = 40;
max_window_size.width = 0;
max_window_size.height = 0;
cvNamedWindow(win_face, 1);
// Carico il file con le informazioni su cosa trovare
cascade = (CvHaarClassifierCascade*)cvLoad(file_xml, 0, 0, 0);
// Alloco la memoria per elaborare i dati
storage = cvCreateMemStorage(0);
/*if(!(capture = cvCaptureFromCAM(wcam)))
{
printf("Impossibile aprire la webcam.\n");
return -1;
}*/
CvSize window_size = cvSize(640, 480);
frame = cvLoadImage("/Users/Gabriele/Desktop/jobs.jpeg");
frame_resized = cvCreateImage(window_size, frame->depth, 3);
frame_resized2 = cvCreateImage(window_size, frame->depth, 3);
cvResize(frame, frame_resized);
CLODEnvironmentData* data = clodInitEnvironment(0);
clifInitBuffers(data->clif, frame_resized->width, frame_resized->height, frame_resized->widthStep, 3);
clodInitBuffers(data, &window_size);
ElapseTime t;;
/* Test grayscale */
IplImage* grayscale = cvCreateImage(cvSize(frame_resized->width, frame_resized->height), IPL_DEPTH_8U, 1);
cvCvtColor(frame_resized, grayscale, CV_BGR2GRAY);
CvMat* sim = cvCreateMat(frame_resized->height + 1, frame_resized->width + 1, CV_32SC1);
CvMat* sqim = cvCreateMat(frame_resized->height + 1, frame_resized->width + 1, CV_64FC1);
cvIntegral(grayscale, sim, sqim);
double temp = sqim->data.db[2000];
CLIFIntegralResult r = clifIntegral(frame_resized, data->clif, CL_TRUE);
cl_ulong temp2 = ((unsigned long*)r.square_image->data.db)[2000];
cvCopyImage(frame_resized, frame_resized2);
t.start();
find_faces_rect_opencv(frame_resized2, min_window_size, max_window_size);
printf("OpenCV: %8.4f ms\n", t.get());
cvShowImage("Sample OpenCV", frame_resized2);
cvCopyImage(frame_resized, frame_resized2);
t.start();
find_faces_rect_opencl(frame_resized2, data, min_window_size, max_window_size, CLOD_PER_STAGE_ITERATIONS | CLOD_PRECOMPUTE_FEATURES, CL_FALSE);
printf("OpenCL (optimized): %8.4f ms\n", t.get());
cvShowImage("Sample OpenCL (optimized)", frame_resized2);
cvCopyImage(frame_resized, frame_resized2);
t.start();
find_faces_rect_opencl(frame_resized2, data, min_window_size, max_window_size, CLOD_PRECOMPUTE_FEATURES, CL_TRUE);
printf(" %8.4f ms (block)\n", t.get());
cvShowImage("Sample OpenCL (optimized, block)", frame_resized2);
cvCopyImage(frame_resized, frame_resized2);
t.start();
find_faces_rect_opencl(frame_resized2, data, min_window_size, max_window_size, CLOD_PRECOMPUTE_FEATURES | CLOD_PER_STAGE_ITERATIONS, CL_FALSE);
printf("OpenCL (per-stage): %8.4f ms\n", t.get());
cvShowImage("Sample OpenCL (per-stage)", frame_resized2);
cvCopyImage(frame_resized, frame_resized2);
t.start();
find_faces_rect_opencl(frame_resized2, data, min_window_size, max_window_size, CLOD_PRECOMPUTE_FEATURES | CLOD_PER_STAGE_ITERATIONS, CL_TRUE);
printf(" %8.4f ms (block)\n", t.get());
cvShowImage("Sample OpenCL (per-stage, block)", frame_resized2);
//frame_resized->imageData =
//printf("OpenCL (per-stage, optimized): %8.4f ms\n", t.get());
//cvShowImage("Sample OpenCL (per-stage, optimized)", frame2);
//detect_faces(frame, &data);
/*
while(1)
{
// Cerco le facce
char fps[1024] = { 0 };
sprintf(fps, "FPS: %4.2f", (1000.0) / (double)(end - start));
CvFont font;
cvInitFont(&font, CV_FONT_HERSHEY_PLAIN, 1.0f, 1.0f);
CvPoint point;
point.x = 10;
point.y = frame->height - 10;
CvScalar scalar = { 255,255,255,1 };
cvPutText(frame, fps, point, &font, scalar);
cvShowImage(win_face, frame);
frame = cvQueryFrame(capture);
if( (cvWaitKey(10) & 255) == 27 ) break;
}
*/
clodReleaseBuffers(data);
clodReleaseEnvironment(data);
free(data);
cvReleaseImage(&frame);
cvReleaseImage(&frame_resized);
cvReleaseCapture(&capture);
cvDestroyWindow(win_face);
return 0;
}
