Mercurial > cgi-bin > hgwebdir.cgi > PR > Applications > VSs > VSs__H264__App

annotate libavcodec/h264_dsp.c @ 2:897f711a7157

find changesets by author, revision, files, or words in the commit message
rearrange to work with autoconf
author Nina Engelhardt <nengel@mailbox.tu-berlin.de>
date Tue, 25 Sep 2012 15:55:33 +0200
parents
children
rev   line source
nengel@2 1 /*
nengel@2 2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
nengel@2 3 * Copyright (c) 2003-2010 Michael Niedermayer <michaelni@gmx.at>
nengel@2 4 *
nengel@2 5 * This file is part of FFmpeg.
nengel@2 6 *
nengel@2 7 * FFmpeg is free software; you can redistribute it and/or
nengel@2 8 * modify it under the terms of the GNU Lesser General Public
nengel@2 9 * License as published by the Free Software Foundation; either
nengel@2 10 * version 2.1 of the License, or (at your option) any later version.
nengel@2 11 *
nengel@2 12 * FFmpeg is distributed in the hope that it will be useful,
nengel@2 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
nengel@2 14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
nengel@2 15 * Lesser General Public License for more details.
nengel@2 16 *
nengel@2 17 * You should have received a copy of the GNU Lesser General Public
nengel@2 18 * License along with FFmpeg; if not, write to the Free Software
nengel@2 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
nengel@2 20 */
nengel@2 21
nengel@2 22 /**
nengel@2 23 * @file
nengel@2 24 * H.264 / AVC / MPEG4 part10 DSP functions.
nengel@2 25 * @author Michael Niedermayer <michaelni@gmx.at>
nengel@2 26 */
nengel@2 27
nengel@2 28 #include <stdint.h>
nengel@2 29 #include "avcodec.h"
nengel@2 30 #include "h264_dsp.h"
nengel@2 31
nengel@2 32 #define op_scale1(x) block[x] = av_clip_uint8( (block[x]*weight + offset) >> log2_denom )
nengel@2 33 #define op_scale2(x) dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
nengel@2 34 #define H264_WEIGHT(W,H) \
nengel@2 35 static void weight_h264_pixels ## W ## x ## H ## _c(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \
nengel@2 36 int y; \
nengel@2 37 offset <<= log2_denom; \
nengel@2 38 if(log2_denom) offset += 1<<(log2_denom-1); \
nengel@2 39 for(y=0; y<H; y++, block += stride){ \
nengel@2 40 op_scale1(0); \
nengel@2 41 op_scale1(1); \
nengel@2 42 if(W==2) continue; \
nengel@2 43 op_scale1(2); \
nengel@2 44 op_scale1(3); \
nengel@2 45 if(W==4) continue; \
nengel@2 46 op_scale1(4); \
nengel@2 47 op_scale1(5); \
nengel@2 48 op_scale1(6); \
nengel@2 49 op_scale1(7); \
nengel@2 50 if(W==8) continue; \
nengel@2 51 op_scale1(8); \
nengel@2 52 op_scale1(9); \
nengel@2 53 op_scale1(10); \
nengel@2 54 op_scale1(11); \
nengel@2 55 op_scale1(12); \
nengel@2 56 op_scale1(13); \
nengel@2 57 op_scale1(14); \
nengel@2 58 op_scale1(15); \
nengel@2 59 } \
nengel@2 60 } \
nengel@2 61 static void biweight_h264_pixels ## W ## x ## H ## _c(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset){ \
nengel@2 62 int y; \
nengel@2 63 offset = ((offset + 1) | 1) << log2_denom; \
nengel@2 64 for(y=0; y<H; y++, dst += stride, src += stride){ \
nengel@2 65 op_scale2(0); \
nengel@2 66 op_scale2(1); \
nengel@2 67 if(W==2) continue; \
nengel@2 68 op_scale2(2); \
nengel@2 69 op_scale2(3); \
nengel@2 70 if(W==4) continue; \
nengel@2 71 op_scale2(4); \
nengel@2 72 op_scale2(5); \
nengel@2 73 op_scale2(6); \
nengel@2 74 op_scale2(7); \
nengel@2 75 if(W==8) continue; \
nengel@2 76 op_scale2(8); \
nengel@2 77 op_scale2(9); \
nengel@2 78 op_scale2(10); \
nengel@2 79 op_scale2(11); \
nengel@2 80 op_scale2(12); \
nengel@2 81 op_scale2(13); \
nengel@2 82 op_scale2(14); \
nengel@2 83 op_scale2(15); \
nengel@2 84 } \
nengel@2 85 }
nengel@2 86
nengel@2 87 H264_WEIGHT(16,16)
nengel@2 88 H264_WEIGHT(16,8)
nengel@2 89 H264_WEIGHT(8,16)
nengel@2 90 H264_WEIGHT(8,8)
nengel@2 91 H264_WEIGHT(8,4)
nengel@2 92 H264_WEIGHT(4,8)
nengel@2 93 H264_WEIGHT(4,4)
nengel@2 94 H264_WEIGHT(4,2)
nengel@2 95 H264_WEIGHT(2,4)
nengel@2 96 H264_WEIGHT(2,2)
nengel@2 97
nengel@2 98 #undef op_scale1
nengel@2 99 #undef op_scale2
nengel@2 100 #undef H264_WEIGHT
nengel@2 101
nengel@2 102 static av_always_inline void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
nengel@2 103 {
nengel@2 104 int i, d;
nengel@2 105 for( i = 0; i < 4; i++ ) {
nengel@2 106 if( tc0[i] < 0 ) {
nengel@2 107 pix += 4*ystride;
nengel@2 108 continue;
nengel@2 109 }
nengel@2 110 for( d = 0; d < 4; d++ ) {
nengel@2 111 const int p0 = pix[-1*xstride];
nengel@2 112 const int p1 = pix[-2*xstride];
nengel@2 113 const int p2 = pix[-3*xstride];
nengel@2 114 const int q0 = pix[0];
nengel@2 115 const int q1 = pix[1*xstride];
nengel@2 116 const int q2 = pix[2*xstride];
nengel@2 117
nengel@2 118 if( FFABS( p0 - q0 ) < alpha &&
nengel@2 119 FFABS( p1 - p0 ) < beta &&
nengel@2 120 FFABS( q1 - q0 ) < beta ) {
nengel@2 121
nengel@2 122 int tc = tc0[i];
nengel@2 123 int i_delta;
nengel@2 124
nengel@2 125 if( FFABS( p2 - p0 ) < beta ) {
nengel@2 126 if(tc0[i])
nengel@2 127 pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
nengel@2 128 tc++;
nengel@2 129 }
nengel@2 130 if( FFABS( q2 - q0 ) < beta ) {
nengel@2 131 if(tc0[i])
nengel@2 132 pix[ xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
nengel@2 133 tc++;
nengel@2 134 }
nengel@2 135
nengel@2 136 i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
nengel@2 137 pix[-xstride] = av_clip_uint8( p0 + i_delta ); /* p0' */
nengel@2 138 pix[0] = av_clip_uint8( q0 - i_delta ); /* q0' */
nengel@2 139 }
nengel@2 140 pix += ystride;
nengel@2 141 }
nengel@2 142 }
nengel@2 143 }
nengel@2 144 static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2 145 {
nengel@2 146 h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0);
nengel@2 147 }
nengel@2 148 static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2 149 {
nengel@2 150 h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
nengel@2 151 }
nengel@2 152
nengel@2 153 static av_always_inline void h264_loop_filter_luma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
nengel@2 154 {
nengel@2 155 int d;
nengel@2 156 for( d = 0; d < 16; d++ ) {
nengel@2 157 const int p2 = pix[-3*xstride];
nengel@2 158 const int p1 = pix[-2*xstride];
nengel@2 159 const int p0 = pix[-1*xstride];
nengel@2 160
nengel@2 161 const int q0 = pix[ 0*xstride];
nengel@2 162 const int q1 = pix[ 1*xstride];
nengel@2 163 const int q2 = pix[ 2*xstride];
nengel@2 164
nengel@2 165 if( FFABS( p0 - q0 ) < alpha &&
nengel@2 166 FFABS( p1 - p0 ) < beta &&
nengel@2 167 FFABS( q1 - q0 ) < beta ) {
nengel@2 168
nengel@2 169 if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
nengel@2 170 if( FFABS( p2 - p0 ) < beta)
nengel@2 171 {
nengel@2 172 const int p3 = pix[-4*xstride];
nengel@2 173 /* p0', p1', p2' */
nengel@2 174 pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
nengel@2 175 pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
nengel@2 176 pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
nengel@2 177 } else {
nengel@2 178 /* p0' */
nengel@2 179 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
nengel@2 180 }
nengel@2 181 if( FFABS( q2 - q0 ) < beta)
nengel@2 182 {
nengel@2 183 const int q3 = pix[3*xstride];
nengel@2 184 /* q0', q1', q2' */
nengel@2 185 pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
nengel@2 186 pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
nengel@2 187 pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
nengel@2 188 } else {
nengel@2 189 /* q0' */
nengel@2 190 pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
nengel@2 191 }
nengel@2 192 }else{
nengel@2 193 /* p0', q0' */
nengel@2 194 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
nengel@2 195 pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
nengel@2 196 }
nengel@2 197 }
nengel@2 198 pix += ystride;
nengel@2 199 }
nengel@2 200 }
nengel@2 201 static void h264_v_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2 202 {
nengel@2 203 h264_loop_filter_luma_intra_c(pix, stride, 1, alpha, beta);
nengel@2 204 }
nengel@2 205 static void h264_h_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2 206 {
nengel@2 207 h264_loop_filter_luma_intra_c(pix, 1, stride, alpha, beta);
nengel@2 208 }
nengel@2 209
nengel@2 210 static av_always_inline void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
nengel@2 211 {
nengel@2 212 int i, d;
nengel@2 213 for( i = 0; i < 4; i++ ) {
nengel@2 214 const int tc = tc0[i];
nengel@2 215 if( tc <= 0 ) {
nengel@2 216 pix += 2*ystride;
nengel@2 217 continue;
nengel@2 218 }
nengel@2 219 for( d = 0; d < 2; d++ ) {
nengel@2 220 const int p0 = pix[-1*xstride];
nengel@2 221 const int p1 = pix[-2*xstride];
nengel@2 222 const int q0 = pix[0];
nengel@2 223 const int q1 = pix[1*xstride];
nengel@2 224
nengel@2 225 if( FFABS( p0 - q0 ) < alpha &&
nengel@2 226 FFABS( p1 - p0 ) < beta &&
nengel@2 227 FFABS( q1 - q0 ) < beta ) {
nengel@2 228
nengel@2 229 int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
nengel@2 230
nengel@2 231 pix[-xstride] = av_clip_uint8( p0 + delta ); /* p0' */
nengel@2 232 pix[0] = av_clip_uint8( q0 - delta ); /* q0' */
nengel@2 233 }
nengel@2 234 pix += ystride;
nengel@2 235 }
nengel@2 236 }
nengel@2 237 }
nengel@2 238 static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2 239 {
nengel@2 240 h264_loop_filter_chroma_c(pix, stride, 1, alpha, beta, tc0);
nengel@2 241 }
nengel@2 242 static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2 243 {
nengel@2 244 h264_loop_filter_chroma_c(pix, 1, stride, alpha, beta, tc0);
nengel@2 245 }
nengel@2 246
nengel@2 247 static av_always_inline void h264_loop_filter_chroma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
nengel@2 248 {
nengel@2 249 int d;
nengel@2 250 for( d = 0; d < 8; d++ ) {
nengel@2 251 const int p0 = pix[-1*xstride];
nengel@2 252 const int p1 = pix[-2*xstride];
nengel@2 253 const int q0 = pix[0];
nengel@2 254 const int q1 = pix[1*xstride];
nengel@2 255
nengel@2 256 if( FFABS( p0 - q0 ) < alpha &&
nengel@2 257 FFABS( p1 - p0 ) < beta &&
nengel@2 258 FFABS( q1 - q0 ) < beta ) {
nengel@2 259
nengel@2 260 pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
nengel@2 261 pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
nengel@2 262 }
nengel@2 263 pix += ystride;
nengel@2 264 }
nengel@2 265 }
nengel@2 266 static void h264_v_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2 267 {
nengel@2 268 h264_loop_filter_chroma_intra_c(pix, stride, 1, alpha, beta);
nengel@2 269 }
nengel@2 270 static void h264_h_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2 271 {
nengel@2 272 h264_loop_filter_chroma_intra_c(pix, 1, stride, alpha, beta);
nengel@2 273 }
nengel@2 274
nengel@2 275 void ff_h264dsp_init(H264DSPContext *c)
nengel@2 276 {
nengel@2 277 c->h264_idct_add= ff_h264_idct_add_c;
nengel@2 278 c->h264_idct8_add= ff_h264_idct8_add_c;
nengel@2 279 c->h264_idct_dc_add= ff_h264_idct_dc_add_c;
nengel@2 280 c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c;
nengel@2 281 c->h264_idct_add16 = ff_h264_idct_add16_c;
nengel@2 282 c->h264_idct8_add4 = ff_h264_idct8_add4_c;
nengel@2 283 c->h264_idct_add8 = ff_h264_idct_add8_c;
nengel@2 284 c->h264_idct_add16intra= ff_h264_idct_add16intra_c;
nengel@2 285
nengel@2 286 c->weight_h264_pixels_tab[0]= weight_h264_pixels16x16_c;
nengel@2 287 c->weight_h264_pixels_tab[1]= weight_h264_pixels16x8_c;
nengel@2 288 c->weight_h264_pixels_tab[2]= weight_h264_pixels8x16_c;
nengel@2 289 c->weight_h264_pixels_tab[3]= weight_h264_pixels8x8_c;
nengel@2 290 c->weight_h264_pixels_tab[4]= weight_h264_pixels8x4_c;
nengel@2 291 c->weight_h264_pixels_tab[5]= weight_h264_pixels4x8_c;
nengel@2 292 c->weight_h264_pixels_tab[6]= weight_h264_pixels4x4_c;
nengel@2 293 c->weight_h264_pixels_tab[7]= weight_h264_pixels4x2_c;
nengel@2 294 c->weight_h264_pixels_tab[8]= weight_h264_pixels2x4_c;
nengel@2 295 c->weight_h264_pixels_tab[9]= weight_h264_pixels2x2_c;
nengel@2 296 c->biweight_h264_pixels_tab[0]= biweight_h264_pixels16x16_c;
nengel@2 297 c->biweight_h264_pixels_tab[1]= biweight_h264_pixels16x8_c;
nengel@2 298 c->biweight_h264_pixels_tab[2]= biweight_h264_pixels8x16_c;
nengel@2 299 c->biweight_h264_pixels_tab[3]= biweight_h264_pixels8x8_c;
nengel@2 300 c->biweight_h264_pixels_tab[4]= biweight_h264_pixels8x4_c;
nengel@2 301 c->biweight_h264_pixels_tab[5]= biweight_h264_pixels4x8_c;
nengel@2 302 c->biweight_h264_pixels_tab[6]= biweight_h264_pixels4x4_c;
nengel@2 303 c->biweight_h264_pixels_tab[7]= biweight_h264_pixels4x2_c;
nengel@2 304 c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c;
nengel@2 305 c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c;
nengel@2 306
nengel@2 307 c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c;
nengel@2 308 c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c;
nengel@2 309 c->h264_v_loop_filter_luma_intra= h264_v_loop_filter_luma_intra_c;
nengel@2 310 c->h264_h_loop_filter_luma_intra= h264_h_loop_filter_luma_intra_c;
nengel@2 311 c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c;
nengel@2 312 c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c;
nengel@2 313 c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c;
nengel@2 314 c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_c;
nengel@2 315 c->h264_loop_filter_strength= NULL;
nengel@2 316
nengel@2 317 if (ARCH_ARM) ff_h264dsp_init_arm(c);
nengel@2 318 if (HAVE_ALTIVEC) ff_h264dsp_init_ppc(c);
nengel@2 319 if (HAVE_MMX) ff_h264dsp_init_x86(c);
nengel@2 320 }