1 /*M///////////////////////////////////////////////////////////////////////////////////////
3 // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
5 // By downloading, copying, installing or using the software you agree to this license.
6 // If you do not agree to this license, do not download, install,
7 // copy or use the software.
11 // For Open Source Computer Vision Library
13 // Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
14 // Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
15 // Third party copyrights are property of their respective owners.
18 // Wang Weiyan, wangweiyanster@gmail.com
19 // Peng Xiao, pengxiao@multicorewareinc.com
21 // Redistribution and use in source and binary forms, with or without modification,
22 // are permitted provided that the following conditions are met:
24 // * Redistribution's of source code must retain the above copyright notice,
25 // this list of conditions and the following disclaimer.
27 // * Redistribution's in binary form must reproduce the above copyright notice,
28 // this list of conditions and the following disclaimer in the documentation
29 // and/or other materials provided with the distribution.
31 // * The name of the copyright holders may not be used to endorse or promote products
32 // derived from this software without specific prior written permission.
34 // This software is provided by the copyright holders and contributors "as is" and
35 // any express or implied warranties, including, but not limited to, the implied
36 // warranties of merchantability and fitness for a particular purpose are disclaimed.
37 // In no event shall the Intel Corporation or contributors be liable for any direct,
38 // indirect, incidental, special, exemplary, or consequential damages
39 // (including, but not limited to, procurement of substitute goods or services;
40 // loss of use, data, or profits; or business interruption) however caused
41 // and on any theory of liability, whether in contract, strict liability,
42 // or tort (including negligence or otherwise) arising in any way out of
43 // the use of this software, even if advised of the possibility of such damage.
47 #include "precomp.hpp"
48 #include "opencl_kernels.hpp"
51 using namespace cv::ocl;
54 #define CV_DESCALE(x, n) (((x) + (1 << ((n)-1))) >> (n))
58 #define FLT_EPSILON 1.192092896e-07F
64 void RGB2Gray_caller(const oclMat &src, oclMat &dst, int bidx)
66 int channels = src.oclchannels();
67 int src_offset = src.offset / src.elemSize1(), src_step = src.step1();
68 int dst_offset = dst.offset / dst.elemSize1(), dst_step = dst.step1();
70 String build_options = format("-D DEPTH_%d", src.depth());
72 std::vector<std::pair<size_t , const void *> > args;
73 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.cols));
74 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.rows));
75 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_step));
76 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step));
77 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&channels));
78 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&bidx));
79 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
80 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
81 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_offset ));
82 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_offset ));
84 size_t gt[3] = {src.cols, src.rows, 1}, lt[3] = {16, 16, 1};
85 openCLExecuteKernel(src.clCxt, &cvt_color, "RGB2Gray", gt, lt, args, -1, -1, build_options.c_str());
88 void Gray2RGB_caller(const oclMat &src, oclMat &dst)
90 String build_options = format("-D DEPTH_%d", src.depth());
91 int src_offset = src.offset / src.elemSize1(), src_step = src.step1();
92 int dst_offset = dst.offset / dst.elemSize1(), dst_step = dst.step1();
94 std::vector<std::pair<size_t , const void *> > args;
95 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.cols));
96 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.rows));
97 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_step));
98 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step));
99 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
100 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
101 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_offset ));
102 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_offset ));
104 size_t gt[3] = {src.cols, src.rows, 1}, lt[3] = {16, 16, 1};
105 openCLExecuteKernel(src.clCxt, &cvt_color, "Gray2RGB", gt, lt, args, -1, -1, build_options.c_str());
108 void RGB2YUV_caller(const oclMat &src, oclMat &dst, int bidx)
110 int channels = src.oclchannels();
111 String build_options = format("-D DEPTH_%d", src.depth());
112 int src_offset = src.offset / src.elemSize1(), src_step = src.step1();
113 int dst_offset = dst.offset / dst.elemSize1(), dst_step = dst.step1();
115 std::vector<std::pair<size_t , const void *> > args;
116 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.cols));
117 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.rows));
118 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_step));
119 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step));
120 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&channels));
121 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&bidx));
122 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
123 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
124 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_offset ));
125 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_offset ));
127 size_t gt[3] = {src.cols, src.rows, 1}, lt[3] = {16, 16, 1};
128 openCLExecuteKernel(src.clCxt, &cvt_color, "RGB2YUV", gt, lt, args, -1, -1, build_options.c_str());
131 void YUV2RGB_caller(const oclMat &src, oclMat &dst, int bidx)
133 int channels = src.oclchannels();
134 int src_offset = src.offset / src.elemSize1(), src_step = src.step1();
135 int dst_offset = dst.offset / dst.elemSize1(), dst_step = dst.step1();
137 String buildOptions = format("-D DEPTH_%d", src.depth());
139 std::vector<std::pair<size_t , const void *> > args;
140 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.cols));
141 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.rows));
142 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_step));
143 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step));
144 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&channels));
145 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&bidx));
146 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
147 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
148 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_offset ));
149 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_offset ));
151 size_t gt[3] = {src.cols, src.rows, 1}, lt[3] = {16, 16, 1};
152 openCLExecuteKernel(src.clCxt, &cvt_color, "YUV2RGB", gt, lt, args, -1, -1, buildOptions.c_str());
155 void YUV2RGB_NV12_caller(const oclMat &src, oclMat &dst, int bidx)
157 String build_options = format("-D DEPTH_%d", src.depth());
158 int src_offset = src.offset / src.elemSize1(), src_step = src.step1();
159 int dst_offset = dst.offset / dst.elemSize1(), dst_step = dst.step1();
161 std::vector<std::pair<size_t , const void *> > args;
162 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.cols));
163 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.rows));
164 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_step));
165 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step));
166 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&bidx));
167 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.cols));
168 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst.rows));
169 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
170 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
171 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_offset ));
172 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_offset ));
174 size_t gt[3] = {dst.cols / 2, dst.rows / 2, 1}, lt[3] = {16, 16, 1};
175 openCLExecuteKernel(src.clCxt, &cvt_color, "YUV2RGBA_NV12", gt, lt, args, -1, -1, build_options.c_str());
178 void RGB2YCrCb_caller(const oclMat &src, oclMat &dst, int bidx)
180 int channels = src.oclchannels();
181 String build_options = format("-D DEPTH_%d", src.depth());
182 int src_offset = src.offset / src.elemSize1(), src_step = src.step1();
183 int dst_offset = dst.offset / dst.elemSize1(), dst_step = dst.step1();
185 std::vector<std::pair<size_t , const void *> > args;
186 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.cols));
187 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src.rows));
188 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_step));
189 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_step));
190 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&channels));
191 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&bidx));
192 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&src.data));
193 args.push_back( std::make_pair( sizeof(cl_mem) , (void *)&dst.data));
194 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&src_offset ));
195 args.push_back( std::make_pair( sizeof(cl_int) , (void *)&dst_offset ));
197 size_t gt[3] = {src.cols, src.rows, 1}, lt[3] = {16, 16, 1};
198 openCLExecuteKernel(src.clCxt, &cvt_color, "RGB2YCrCb", gt, lt, args, -1, -1, build_options.c_str());
201 void cvtColor_caller(const oclMat &src, oclMat &dst, int code, int dcn)
203 Size sz = src.size();
204 int scn = src.oclchannels(), depth = src.depth(), bidx;
206 CV_Assert(depth == CV_8U || depth == CV_16U || depth == CV_32F);
211 case COLOR_BGR2BGRA: case COLOR_RGB2BGRA: case COLOR_BGRA2BGR:
212 case COLOR_RGBA2BGR: case COLOR_RGB2BGR: case COLOR_BGRA2RGBA:
213 case COLOR_BGR2BGR565: case COLOR_BGR2BGR555: case COLOR_RGB2BGR565: case COLOR_RGB2BGR555:
214 case COLOR_BGRA2BGR565: case COLOR_BGRA2BGR555: case COLOR_RGBA2BGR565: case COLOR_RGBA2BGR555:
215 case COLOR_BGR5652BGR: case COLOR_BGR5552BGR: case COLOR_BGR5652RGB: case COLOR_BGR5552RGB:
216 case COLOR_BGR5652BGRA: case COLOR_BGR5552BGRA: case COLOR_BGR5652RGBA: case COLOR_BGR5552RGBA:
219 case COLOR_BGRA2GRAY:
221 case COLOR_RGBA2GRAY:
223 CV_Assert(scn == 3 || scn == 4);
224 bidx = code == COLOR_BGR2GRAY || code == COLOR_BGRA2GRAY ? 0 : 2;
225 dst.create(sz, CV_MAKETYPE(depth, 1));
226 RGB2Gray_caller(src, dst, bidx);
230 case COLOR_GRAY2BGRA:
233 dcn = code == COLOR_GRAY2BGRA ? 4 : 3;
234 dst.create(sz, CV_MAKETYPE(depth, dcn));
235 Gray2RGB_caller(src, dst);
241 CV_Assert(scn == 3 || scn == 4);
242 bidx = code == COLOR_RGB2YUV ? 0 : 2;
243 dst.create(sz, CV_MAKETYPE(depth, 3));
244 RGB2YUV_caller(src, dst, bidx);
250 CV_Assert(scn == 3 || scn == 4);
251 bidx = code == COLOR_YUV2RGB ? 0 : 2;
252 dst.create(sz, CV_MAKETYPE(depth, 3));
253 YUV2RGB_caller(src, dst, bidx);
256 case COLOR_YUV2RGB_NV12:
257 case COLOR_YUV2BGR_NV12:
258 case COLOR_YUV2RGBA_NV12:
259 case COLOR_YUV2BGRA_NV12:
262 CV_Assert( sz.width % 2 == 0 && sz.height % 3 == 0 && depth == CV_8U );
263 dcn = code == COLOR_YUV2BGRA_NV12 || code == COLOR_YUV2RGBA_NV12 ? 4 : 3;
264 bidx = code == COLOR_YUV2BGRA_NV12 || code == COLOR_YUV2BGR_NV12 ? 0 : 2;
266 Size dstSz(sz.width, sz.height * 2 / 3);
267 dst.create(dstSz, CV_MAKETYPE(depth, dcn));
268 YUV2RGB_NV12_caller(src, dst, bidx);
271 case COLOR_BGR2YCrCb:
272 case COLOR_RGB2YCrCb:
274 CV_Assert(scn == 3 || scn == 4);
275 bidx = code == COLOR_BGR2YCrCb ? 0 : 2;
276 dst.create(sz, CV_MAKETYPE(depth, 3));
277 RGB2YCrCb_caller(src, dst, bidx);
280 case COLOR_YCrCb2BGR:
281 case COLOR_YCrCb2RGB:
286 case COLOR_BGR5652GRAY: case COLOR_BGR5552GRAY:
287 case COLOR_GRAY2BGR565: case COLOR_GRAY2BGR555:
288 case COLOR_BGR2YCrCb: case COLOR_RGB2YCrCb:
289 case COLOR_BGR2XYZ: case COLOR_RGB2XYZ:
290 case COLOR_XYZ2BGR: case COLOR_XYZ2RGB:
291 case COLOR_BGR2HSV: case COLOR_RGB2HSV: case COLOR_BGR2HSV_FULL: case COLOR_RGB2HSV_FULL:
292 case COLOR_BGR2HLS: case COLOR_RGB2HLS: case COLOR_BGR2HLS_FULL: case COLOR_RGB2HLS_FULL:
293 case COLOR_HSV2BGR: case COLOR_HSV2RGB: case COLOR_HSV2BGR_FULL: case COLOR_HSV2RGB_FULL:
294 case COLOR_HLS2BGR: case COLOR_HLS2RGB: case COLOR_HLS2BGR_FULL: case COLOR_HLS2RGB_FULL:
297 CV_Error(Error::StsBadFlag, "Unknown/unsupported color conversion code" );
302 void cv::ocl::cvtColor(const oclMat &src, oclMat &dst, int code, int dcn)
304 cvtColor_caller(src, dst, code, dcn);