1// License Agreement 2// For Open Source Computer Vision Library 3// 4// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved. 5// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved. 6// Third party copyrights are property of their respective owners. 7// 8// @Authors 9// Niko Li, newlife20080214@gmail.com 10// Zero Lin, zero.lin@amd.com 11// Yao Wang, bitwangyaoyao@gmail.com 12// Redistribution and use in source and binary forms, with or without modification, 13// are permitted provided that the following conditions are met: 14// 15// * Redistribution's of source code must retain the above copyright notice, 16// this list of conditions and the following disclaimer. 17// 18// * Redistribution's in binary form must reproduce the above copyright notice, 19// this list of conditions and the following disclaimer in the documentation 20// and/or other materials provided with the distribution. 21// 22// * The name of the copyright holders may not be used to endorse or promote products 23// derived from this software without specific prior written permission. 24// 25// This software is provided by the copyright holders and contributors as is and 26// any express or implied warranties, including, but not limited to, the implied 27// warranties of merchantability and fitness for a particular purpose are disclaimed. 28// In no event shall the Intel Corporation or contributors be liable for any direct, 29// indirect, incidental, special, exemplary, or consequential damages 30// (including, but not limited to, procurement of substitute goods or services; 31// loss of use, data, or profits; or business interruption) however caused 32// and on any theory of liability, whether in contract, strict liability, 33// or tort (including negligence or otherwise) arising in any way out of 34// the use of this software, even if advised of the possibility of such damage. 35// 36// 37 38#ifdef DOUBLE_SUPPORT 39#ifdef cl_amd_fp64 40#pragma OPENCL EXTENSION cl_amd_fp64:enable 41#elif defined (cl_khr_fp64) 42#pragma OPENCL EXTENSION cl_khr_fp64:enable 43#endif 44#endif 45 46#define noconvert 47 48#if cn != 3 49#define loadpix(addr) *(__global const T *)(addr) 50#define storepix(val, addr) *(__global T *)(addr) = val 51#define TSIZE (int)sizeof(T) 52#else 53#define loadpix(addr) vload3(0, (__global const T1 *)(addr)) 54#define storepix(val, addr) vstore3(val, 0, (__global T1 *)(addr)) 55#define TSIZE ((int)sizeof(T1)*3) 56#endif 57 58#ifdef DEPTH_0 59#define MIN_VAL 0 60#define MAX_VAL UCHAR_MAX 61#elif defined DEPTH_1 62#define MIN_VAL SCHAR_MIN 63#define MAX_VAL SCHAR_MAX 64#elif defined DEPTH_2 65#define MIN_VAL 0 66#define MAX_VAL USHRT_MAX 67#elif defined DEPTH_3 68#define MIN_VAL SHRT_MIN 69#define MAX_VAL SHRT_MAX 70#elif defined DEPTH_4 71#define MIN_VAL INT_MIN 72#define MAX_VAL INT_MAX 73#elif defined DEPTH_5 74#define MIN_VAL (-FLT_MAX) 75#define MAX_VAL FLT_MAX 76#elif defined DEPTH_6 77#define MIN_VAL (-DBL_MAX) 78#define MAX_VAL DBL_MAX 79#endif 80 81#ifdef OP_ERODE 82#define VAL MAX_VAL 83#elif defined OP_DILATE 84#define VAL MIN_VAL 85#else 86#error "Unknown operation" 87#endif 88 89#ifdef OP_ERODE 90#if defined INTEL_DEVICE && defined DEPTH_0 91#define MORPH_OP(A, B) ((A) < (B) ? (A) : (B)) 92#else 93#define MORPH_OP(A, B) min((A), (B)) 94#endif 95#endif 96#ifdef OP_DILATE 97#define MORPH_OP(A, B) max((A), (B)) 98#endif 99 100#define PROCESS(y, x) \ 101 temp = LDS_DAT[mad24(l_y + y, width, l_x + x)]; \ 102 res = MORPH_OP(res, temp); 103 104// BORDER_CONSTANT: iiiiii|abcdefgh|iiiiiii 105#define ELEM(i, l_edge, r_edge, elem1, elem2) (i) < (l_edge) | (i) >= (r_edge) ? (elem1) : (elem2) 106 107#if defined OP_GRADIENT || defined OP_TOPHAT || defined OP_BLACKHAT 108#define EXTRA_PARAMS , __global const uchar * matptr, int mat_step, int mat_offset 109#else 110#define EXTRA_PARAMS 111#endif 112 113__kernel void morph(__global const uchar * srcptr, int src_step, int src_offset, 114 __global uchar * dstptr, int dst_step, int dst_offset, 115 int src_offset_x, int src_offset_y, int cols, int rows, 116 int src_whole_cols, int src_whole_rows EXTRA_PARAMS) 117{ 118 int gidx = get_global_id(0), gidy = get_global_id(1); 119 int l_x = get_local_id(0), l_y = get_local_id(1); 120 int x = get_group_id(0) * LSIZE0, y = get_group_id(1) * LSIZE1; 121 int start_x = x + src_offset_x - RADIUSX; 122 int width = mad24(RADIUSX, 2, LSIZE0 + 1); 123 int start_y = y + src_offset_y - RADIUSY; 124 int point1 = mad24(l_y, LSIZE0, l_x); 125 int point2 = point1 + LSIZE0 * LSIZE1; 126 int tl_x = point1 % width, tl_y = point1 / width; 127 int tl_x2 = point2 % width, tl_y2 = point2 / width; 128 int cur_x = start_x + tl_x, cur_y = start_y + tl_y; 129 int cur_x2 = start_x + tl_x2, cur_y2 = start_y + tl_y2; 130 int start_addr = mad24(cur_y, src_step, cur_x * TSIZE); 131 int start_addr2 = mad24(cur_y2, src_step, cur_x2 * TSIZE); 132 133 __local T LDS_DAT[2 * LSIZE1 * LSIZE0]; 134 135 // read pixels from src 136 int end_addr = mad24(src_whole_rows - 1, src_step, src_whole_cols * TSIZE); 137 start_addr = start_addr < end_addr && start_addr > 0 ? start_addr : 0; 138 start_addr2 = start_addr2 < end_addr && start_addr2 > 0 ? start_addr2 : 0; 139 140 T temp0 = loadpix(srcptr + start_addr); 141 T temp1 = loadpix(srcptr + start_addr2); 142 143 // judge if read out of boundary 144 temp0 = ELEM(cur_x, 0, src_whole_cols, (T)(VAL), temp0); 145 temp0 = ELEM(cur_y, 0, src_whole_rows, (T)(VAL), temp0); 146 147 temp1 = ELEM(cur_x2, 0, src_whole_cols, (T)(VAL), temp1); 148 temp1 = ELEM(cur_y2, 0, src_whole_rows, (T)(VAL), temp1); 149 150 LDS_DAT[point1] = temp0; 151 LDS_DAT[point2] = temp1; 152 barrier(CLK_LOCAL_MEM_FENCE); 153 154 if (gidx < cols && gidy < rows) 155 { 156 T res = (T)(VAL), temp; 157 PROCESS_ELEMS; 158 159 int dst_index = mad24(gidy, dst_step, mad24(gidx, TSIZE, dst_offset)); 160 161#if defined OP_GRADIENT || defined OP_TOPHAT || defined OP_BLACKHAT 162 int mat_index = mad24(gidy, mat_step, mad24(gidx, TSIZE, mat_offset)); 163 T value = loadpix(matptr + mat_index); 164 165#ifdef OP_GRADIENT 166 storepix(convertToT(convertToWT(res) - convertToWT(value)), dstptr + dst_index); 167#elif defined OP_TOPHAT 168 storepix(convertToT(convertToWT(value) - convertToWT(res)), dstptr + dst_index); 169#elif defined OP_BLACKHAT 170 storepix(convertToT(convertToWT(res) - convertToWT(value)), dstptr + dst_index); 171#endif 172#else // erode or dilate 173 storepix(res, dstptr + dst_index); 174#endif 175 } 176} 177