third_party/opencv
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
6eed90e5e8
opencv/modules/ocl/src/match_template.cpp
Niko 5df77a841e remove redundant OPENCL_DIR flag 
remove as much warnings as possible
use enum instead of MACRO for ocl.hpp
add command line parser in accuracy test and perf test
some bug fix for arthim functions
2012-10-22 15:14:22 +08:00

502 lines
24 KiB
C++
Raw Blame History

/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Peng Xiao, pengxiao@multicorewareinc.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other oclMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include <iomanip>
#include "precomp.hpp"
using namespace cv;
using namespace cv::ocl;
using namespace std;
#if !defined (HAVE_OPENCL)
void cv::ocl::matchTemplate(const oclMat &, const oclMat &, oclMat &)
{
throw_nogpu();
}
#else
//helper routines
namespace cv
{
namespace ocl
{
///////////////////////////OpenCL kernel strings///////////////////////////
extern const char *match_template;
}
}
namespace cv
{
namespace ocl
{
void matchTemplate_SQDIFF(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
void matchTemplate_SQDIFF_NORMED(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
void matchTemplate_CCORR(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
void matchTemplate_CCORR_NORMED(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
void matchTemplate_CCOFF(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
void matchTemplate_CCOFF_NORMED(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf);
void matchTemplateNaive_SQDIFF(
const oclMat &image, const oclMat &templ, oclMat &result, int cn);
void matchTemplateNaive_CCORR(
const oclMat &image, const oclMat &templ, oclMat &result, int cn);
// Evaluates optimal template's area threshold. If
// template's area is less than the threshold, we use naive match
// template version, otherwise FFT-based (if available)
int getTemplateThreshold(int method, int depth)
{
switch (method)
{
case CV_TM_CCORR:
if (depth == CV_32F) return 250;
if (depth == CV_8U) return 300;
break;
case CV_TM_SQDIFF:
if (depth == CV_32F) return 0x7fffffff; // do naive SQDIFF for CV_32F
if (depth == CV_8U) return 300;
break;
}
CV_Error(CV_StsBadArg, "getTemplateThreshold: unsupported match template mode");
return 0;
}
//////////////////////////////////////////////////////////////////////
// SQDIFF
void matchTemplate_SQDIFF(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &)
{
result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
if (templ.size().area() < getTemplateThreshold(CV_TM_SQDIFF, image.depth()))
{
matchTemplateNaive_SQDIFF(image, templ, result, image.oclchannels());
return;
}
else
{
// TODO
CV_Error(CV_StsBadArg, "Not supported yet for this size template");
}
}
void matchTemplate_SQDIFF_NORMED(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf)
{
matchTemplate_CCORR(image, templ, result, buf);
buf.image_sums.resize(1);
integral(image.reshape(1), buf.image_sums[0]);
unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
Context *clCxt = image.clCxt;
string kernelName = "matchTemplate_Prepared_SQDIFF_NORMED";
vector< pair<size_t, const void *> > args;
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[0].data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&result.data));
args.push_back( make_pair( sizeof(cl_ulong), (void *)&templ_sqsum));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].step));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.step));
size_t globalThreads[3] = {result.cols, result.rows, 1};
size_t localThreads[3] = {32, 8, 1};
openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, 1, CV_8U);
}
void matchTemplateNaive_SQDIFF(
const oclMat &image, const oclMat &templ, oclMat &result, int)
{
CV_Assert((image.depth() == CV_8U && templ.depth() == CV_8U )
|| ((image.depth() == CV_32F && templ.depth() == CV_32F) && result.depth() == CV_32F)
);
CV_Assert(image.oclchannels() == templ.oclchannels() && (image.oclchannels() == 1 || image.oclchannels() == 4) && result.oclchannels() == 1);
CV_Assert(result.rows == image.rows - templ.rows + 1 && result.cols == image.cols - templ.cols + 1);
Context *clCxt = image.clCxt;
string kernelName = "matchTemplate_Naive_SQDIFF";
vector< pair<size_t, const void *> > args;
args.push_back( make_pair( sizeof(cl_mem), (void *)&image.data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&templ.data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&result.data));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.step));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.step));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.step));
size_t globalThreads[3] = {result.cols, result.rows, 1};
size_t localThreads[3] = {32, 8, 1};
openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, image.oclchannels(), image.depth());
}
//////////////////////////////////////////////////////////////////////
// CCORR
void matchTemplate_CCORR(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf)
{
result.create(image.rows - templ.rows + 1, image.cols - templ.cols + 1, CV_32F);
if (templ.size().area() < getTemplateThreshold(CV_TM_SQDIFF, image.depth()))
{
matchTemplateNaive_CCORR(image, templ, result, image.oclchannels());
return;
}
else
{
CV_Error(CV_StsBadArg, "Not supported yet for this size template");
if(image.depth() == CV_8U && templ.depth() == CV_8U)
{
image.convertTo(buf.imagef, CV_32F);
templ.convertTo(buf.templf, CV_32F);
}
CV_Assert(image.oclchannels() == 1);
oclMat o_result(image.size(), CV_MAKETYPE(CV_32F, image.oclchannels()));
filter2D(buf.imagef, o_result, CV_32F, buf.templf, Point(0, 0));
result = o_result(Rect(0, 0, image.rows - templ.rows + 1, image.cols - templ.cols + 1));
}
}
void matchTemplate_CCORR_NORMED(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf)
{
matchTemplate_CCORR(image, templ, result, buf);
buf.image_sums.resize(1);
buf.image_sqsums.resize(1);
integral(image.reshape(1), buf.image_sums[0], buf.image_sqsums[0]);
unsigned long long templ_sqsum = (unsigned long long)sqrSum(templ.reshape(1))[0];
Context *clCxt = image.clCxt;
string kernelName = "normalizeKernel";
vector< pair<size_t, const void *> > args;
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[0].data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&result.data));
args.push_back( make_pair( sizeof(cl_ulong), (void *)&templ_sqsum));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].step));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.step));
size_t globalThreads[3] = {result.cols, result.rows, 1};
size_t localThreads[3] = {32, 8, 1};
openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, 1, CV_8U);
}
void matchTemplateNaive_CCORR(
const oclMat &image, const oclMat &templ, oclMat &result, int)
{
CV_Assert((image.depth() == CV_8U && templ.depth() == CV_8U )
|| ((image.depth() == CV_32F && templ.depth() == CV_32F) && result.depth() == CV_32F)
);
CV_Assert(image.oclchannels() == templ.oclchannels() && (image.oclchannels() == 1 || image.oclchannels() == 4) && result.oclchannels() == 1);
CV_Assert(result.rows == image.rows - templ.rows + 1 && result.cols == image.cols - templ.cols + 1);
Context *clCxt = image.clCxt;
string kernelName = "matchTemplate_Naive_CCORR";
vector< pair<size_t, const void *> > args;
args.push_back( make_pair( sizeof(cl_mem), (void *)&image.data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&templ.data));
args.push_back( make_pair( sizeof(cl_mem), (void *)&result.data));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.rows));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.cols));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&image.step));
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.step));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.step));
size_t globalThreads[3] = {result.cols, result.rows, 1};
size_t localThreads[3] = {32, 8, 1};
openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, image.oclchannels(), image.depth());
}
//////////////////////////////////////////////////////////////////////
// CCOFF
void matchTemplate_CCOFF(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf)
{
CV_Assert(image.depth() == CV_8U && templ.depth() == CV_8U);
matchTemplate_CCORR(image, templ, result, buf);
Context *clCxt = image.clCxt;
string kernelName;
kernelName = "matchTemplate_Prepared_CCOFF";
size_t globalThreads[3] = {result.cols, result.rows, 1};
size_t localThreads[3] = {32, 8, 1};
vector< pair<size_t, const void *> > args;
args.push_back( make_pair( sizeof(cl_mem), (void *)&result.data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&image.rows) );
args.push_back( make_pair( sizeof(cl_int), (void *)&image.cols) );
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.rows) );
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.cols) );
args.push_back( make_pair( sizeof(cl_int), (void *)&result.rows) );
args.push_back( make_pair( sizeof(cl_int), (void *)&result.cols) );
args.push_back( make_pair( sizeof(cl_int), (void *)&result.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.step));
// to be continued in the following section
if(image.oclchannels() == 1)
{
buf.image_sums.resize(1);
integral(image, buf.image_sums[0]);
float templ_sum = 0;
templ_sum = (float)sum(templ)[0] / templ.size().area();
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[0].data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].offset) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].step) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum) );
}
else
{
Vec4f templ_sum = Vec4f::all(0);
split(image, buf.images);
templ_sum = sum(templ) / templ.size().area();
buf.image_sums.resize(buf.images.size());
for(int i = 0; i < image.oclchannels(); i ++)
{
integral(buf.images[i], buf.image_sums[i]);
}
switch(image.oclchannels())
{
case 4:
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[0].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[1].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[2].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[3].data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].offset) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].step) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[0]) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[1]) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[2]) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[3]) );
break;
default:
CV_Error(CV_StsBadArg, "matchTemplate: unsupported number of channels");
break;
}
}
openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, image.oclchannels(), image.depth());
}
void matchTemplate_CCOFF_NORMED(
const oclMat &image, const oclMat &templ, oclMat &result, MatchTemplateBuf &buf)
{
image.convertTo(buf.imagef, CV_32F);
templ.convertTo(buf.templf, CV_32F);
matchTemplate_CCORR(buf.imagef, buf.templf, result, buf);
float scale = 1.f / templ.size().area();
Context *clCxt = image.clCxt;
string kernelName;
kernelName = "matchTemplate_Prepared_CCOFF_NORMED";
size_t globalThreads[3] = {result.cols, result.rows, 1};
size_t localThreads[3] = {32, 8, 1};
vector< pair<size_t, const void *> > args;
args.push_back( make_pair( sizeof(cl_mem), (void *)&result.data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&image.rows) );
args.push_back( make_pair( sizeof(cl_int), (void *)&image.cols) );
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.rows) );
args.push_back( make_pair( sizeof(cl_int), (void *)&templ.cols) );
args.push_back( make_pair( sizeof(cl_int), (void *)&result.rows) );
args.push_back( make_pair( sizeof(cl_int), (void *)&result.cols) );
args.push_back( make_pair( sizeof(cl_int), (void *)&result.offset));
args.push_back( make_pair( sizeof(cl_int), (void *)&result.step));
args.push_back( make_pair( sizeof(cl_float), (void *)&scale) );
// to be continued in the following section
if(image.oclchannels() == 1)
{
buf.image_sums.resize(1);
buf.image_sqsums.resize(1);
integral(image, buf.image_sums[0], buf.image_sqsums[0]);
float templ_sum = 0;
float templ_sqsum = 0;
templ_sum = (float)sum(templ)[0];
templ_sqsum = sqrSum(templ)[0];
templ_sqsum -= scale * templ_sum * templ_sum;
templ_sum *= scale;
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[0].data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].offset) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].step) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[0].data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].offset) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].step) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sqsum) );
}
else
{
Vec4f templ_sum = Vec4f::all(0);
Vec4f templ_sqsum = Vec4f::all(0);
split(image, buf.images);
templ_sum = sum(templ);
templ_sqsum = sqrSum(templ);
templ_sqsum -= scale * templ_sum * templ_sum;
float templ_sqsum_sum = 0;
for(int i = 0; i < image.oclchannels(); i ++)
{
templ_sqsum_sum += templ_sqsum[i] - scale * templ_sum[i] * templ_sum[i];
}
templ_sum *= scale;
buf.image_sums.resize(buf.images.size());
buf.image_sqsums.resize(buf.images.size());
for(int i = 0; i < image.oclchannels(); i ++)
{
integral(buf.images[i], buf.image_sums[i], buf.image_sqsums[i]);
}
switch(image.oclchannels())
{
case 4:
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[0].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[1].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[2].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sums[3].data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].offset) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sums[0].step) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[0].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[1].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[2].data) );
args.push_back( make_pair( sizeof(cl_mem), (void *)&buf.image_sqsums[3].data) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].offset) );
args.push_back( make_pair( sizeof(cl_int), (void *)&buf.image_sqsums[0].step) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[0]) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[1]) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[2]) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sum[3]) );
args.push_back( make_pair( sizeof(cl_float), (void *)&templ_sqsum_sum) );
break;
default:
CV_Error(CV_StsBadArg, "matchTemplate: unsupported number of channels");
break;
}
}
openCLExecuteKernel(clCxt, &match_template, kernelName, globalThreads, localThreads, args, image.oclchannels(), image.depth());
}
}/*ocl*/
} /*cv*/
void cv::ocl::matchTemplate(const oclMat &image, const oclMat &templ, oclMat &result, int method)
{
MatchTemplateBuf buf;
matchTemplate(image, templ, result, method, buf);
}
void cv::ocl::matchTemplate(const oclMat &image, const oclMat &templ, oclMat &result, int method, MatchTemplateBuf &buf)
{
CV_Assert(image.type() == templ.type());
CV_Assert(image.cols >= templ.cols && image.rows >= templ.rows);
typedef void (*Caller)(const oclMat &, const oclMat &, oclMat &, MatchTemplateBuf &);
const Caller callers[] =
{
::matchTemplate_SQDIFF, ::matchTemplate_SQDIFF_NORMED,
::matchTemplate_CCORR, ::matchTemplate_CCORR_NORMED,
::matchTemplate_CCOFF, ::matchTemplate_CCOFF_NORMED
};
Caller caller = callers[method];
CV_Assert(caller);
caller(image, templ, result, buf);
}
#endif //
Reference in New Issue View Git Blame Copy Permalink