FFmpeg: libavcodec/h264dsp_template.c Source File

00001 /*
00002  * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
00003  * Copyright (c) 2003-2011 Michael Niedermayer <michaelni@gmx.at>
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00028 #include "high_bit_depth.h"
00029 
00030 #define op_scale1(x)  block[x] = av_clip_pixel( (block[x]*weight + offset) >> log2_denom )
00031 #define op_scale2(x)  dst[x] = av_clip_pixel( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
00032 #define H264_WEIGHT(W,H) \
00033 static void FUNCC(weight_h264_pixels ## W ## x ## H)(uint8_t *p_block, int stride, int log2_denom, int weight, int offset){ \
00034     int y; \
00035     pixel *block = (pixel*)p_block; \
00036     stride >>= sizeof(pixel)-1; \
00037     offset <<= (log2_denom + (BIT_DEPTH-8)); \
00038     if(log2_denom) offset += 1<<(log2_denom-1); \
00039     for(y=0; y<H; y++, block += stride){ \
00040         op_scale1(0); \
00041         op_scale1(1); \
00042         if(W==2) continue; \
00043         op_scale1(2); \
00044         op_scale1(3); \
00045         if(W==4) continue; \
00046         op_scale1(4); \
00047         op_scale1(5); \
00048         op_scale1(6); \
00049         op_scale1(7); \
00050         if(W==8) continue; \
00051         op_scale1(8); \
00052         op_scale1(9); \
00053         op_scale1(10); \
00054         op_scale1(11); \
00055         op_scale1(12); \
00056         op_scale1(13); \
00057         op_scale1(14); \
00058         op_scale1(15); \
00059     } \
00060 } \
00061 static void FUNCC(biweight_h264_pixels ## W ## x ## H)(uint8_t *_dst, uint8_t *_src, int stride, int log2_denom, int weightd, int weights, int offset){ \
00062     int y; \
00063     pixel *dst = (pixel*)_dst; \
00064     pixel *src = (pixel*)_src; \
00065     stride >>= sizeof(pixel)-1; \
00066     offset <<= (BIT_DEPTH-8); \
00067     offset = ((offset + 1) | 1) << log2_denom; \
00068     for(y=0; y<H; y++, dst += stride, src += stride){ \
00069         op_scale2(0); \
00070         op_scale2(1); \
00071         if(W==2) continue; \
00072         op_scale2(2); \
00073         op_scale2(3); \
00074         if(W==4) continue; \
00075         op_scale2(4); \
00076         op_scale2(5); \
00077         op_scale2(6); \
00078         op_scale2(7); \
00079         if(W==8) continue; \
00080         op_scale2(8); \
00081         op_scale2(9); \
00082         op_scale2(10); \
00083         op_scale2(11); \
00084         op_scale2(12); \
00085         op_scale2(13); \
00086         op_scale2(14); \
00087         op_scale2(15); \
00088     } \
00089 }
00090 
00091 H264_WEIGHT(16,16)
00092 H264_WEIGHT(16,8)
00093 H264_WEIGHT(8,16)
00094 H264_WEIGHT(8,8)
00095 H264_WEIGHT(8,4)
00096 H264_WEIGHT(4,8)
00097 H264_WEIGHT(4,4)
00098 H264_WEIGHT(4,2)
00099 H264_WEIGHT(2,4)
00100 H264_WEIGHT(2,2)
00101 
00102 #undef op_scale1
00103 #undef op_scale2
00104 #undef H264_WEIGHT
00105 
00106 static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0)
00107 {
00108     pixel *pix = (pixel*)p_pix;
00109     int i, d;
00110     xstride >>= sizeof(pixel)-1;
00111     ystride >>= sizeof(pixel)-1;
00112     alpha <<= BIT_DEPTH - 8;
00113     beta  <<= BIT_DEPTH - 8;
00114     for( i = 0; i < 4; i++ ) {
00115         const int tc_orig = tc0[i] << (BIT_DEPTH - 8);
00116         if( tc_orig < 0 ) {
00117             pix += inner_iters*ystride;
00118             continue;
00119         }
00120         for( d = 0; d < inner_iters; d++ ) {
00121             const int p0 = pix[-1*xstride];
00122             const int p1 = pix[-2*xstride];
00123             const int p2 = pix[-3*xstride];
00124             const int q0 = pix[0];
00125             const int q1 = pix[1*xstride];
00126             const int q2 = pix[2*xstride];
00127 
00128             if( FFABS( p0 - q0 ) < alpha &&
00129                 FFABS( p1 - p0 ) < beta &&
00130                 FFABS( q1 - q0 ) < beta ) {
00131 
00132                 int tc = tc_orig;
00133                 int i_delta;
00134 
00135                 if( FFABS( p2 - p0 ) < beta ) {
00136                     if(tc_orig)
00137                     pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc_orig, tc_orig );
00138                     tc++;
00139                 }
00140                 if( FFABS( q2 - q0 ) < beta ) {
00141                     if(tc_orig)
00142                     pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc_orig, tc_orig );
00143                     tc++;
00144                 }
00145 
00146                 i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
00147                 pix[-xstride] = av_clip_pixel( p0 + i_delta );    /* p0' */
00148                 pix[0]        = av_clip_pixel( q0 - i_delta );    /* q0' */
00149             }
00150             pix += ystride;
00151         }
00152     }
00153 }
00154 static void FUNCC(h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
00155 {
00156     FUNCC(h264_loop_filter_luma)(pix, stride, sizeof(pixel), 4, alpha, beta, tc0);
00157 }
00158 static void FUNCC(h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
00159 {
00160     FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0);
00161 }
00162 static void FUNCC(h264_h_loop_filter_luma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
00163 {
00164     FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
00165 }
00166 
00167 static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma_intra)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta)
00168 {
00169     pixel *pix = (pixel*)p_pix;
00170     int d;
00171     xstride >>= sizeof(pixel)-1;
00172     ystride >>= sizeof(pixel)-1;
00173     alpha <<= BIT_DEPTH - 8;
00174     beta  <<= BIT_DEPTH - 8;
00175     for( d = 0; d < 4 * inner_iters; d++ ) {
00176         const int p2 = pix[-3*xstride];
00177         const int p1 = pix[-2*xstride];
00178         const int p0 = pix[-1*xstride];
00179 
00180         const int q0 = pix[ 0*xstride];
00181         const int q1 = pix[ 1*xstride];
00182         const int q2 = pix[ 2*xstride];
00183 
00184         if( FFABS( p0 - q0 ) < alpha &&
00185             FFABS( p1 - p0 ) < beta &&
00186             FFABS( q1 - q0 ) < beta ) {
00187 
00188             if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
00189                 if( FFABS( p2 - p0 ) < beta)
00190                 {
00191                     const int p3 = pix[-4*xstride];
00192                     /* p0', p1', p2' */
00193                     pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
00194                     pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
00195                     pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
00196                 } else {
00197                     /* p0' */
00198                     pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
00199                 }
00200                 if( FFABS( q2 - q0 ) < beta)
00201                 {
00202                     const int q3 = pix[3*xstride];
00203                     /* q0', q1', q2' */
00204                     pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
00205                     pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
00206                     pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
00207                 } else {
00208                     /* q0' */
00209                     pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
00210                 }
00211             }else{
00212                 /* p0', q0' */
00213                 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
00214                 pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
00215             }
00216         }
00217         pix += ystride;
00218     }
00219 }
00220 static void FUNCC(h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
00221 {
00222     FUNCC(h264_loop_filter_luma_intra)(pix, stride, sizeof(pixel), 4, alpha, beta);
00223 }
00224 static void FUNCC(h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
00225 {
00226     FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta);
00227 }
00228 static void FUNCC(h264_h_loop_filter_luma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
00229 {
00230     FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
00231 }
00232 
00233 static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0)
00234 {
00235     pixel *pix = (pixel*)p_pix;
00236     int i, d;
00237     alpha <<= BIT_DEPTH - 8;
00238     beta  <<= BIT_DEPTH - 8;
00239     xstride >>= sizeof(pixel)-1;
00240     ystride >>= sizeof(pixel)-1;
00241     for( i = 0; i < 4; i++ ) {
00242         const int tc = ((tc0[i] - 1) << (BIT_DEPTH - 8)) + 1;
00243         if( tc <= 0 ) {
00244             pix += inner_iters*ystride;
00245             continue;
00246         }
00247         for( d = 0; d < inner_iters; d++ ) {
00248             const int p0 = pix[-1*xstride];
00249             const int p1 = pix[-2*xstride];
00250             const int q0 = pix[0];
00251             const int q1 = pix[1*xstride];
00252 
00253             if( FFABS( p0 - q0 ) < alpha &&
00254                 FFABS( p1 - p0 ) < beta &&
00255                 FFABS( q1 - q0 ) < beta ) {
00256 
00257                 int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
00258 
00259                 pix[-xstride] = av_clip_pixel( p0 + delta );    /* p0' */
00260                 pix[0]        = av_clip_pixel( q0 - delta );    /* q0' */
00261             }
00262             pix += ystride;
00263         }
00264     }
00265 }
00266 static void FUNCC(h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
00267 {
00268     FUNCC(h264_loop_filter_chroma)(pix, stride, sizeof(pixel), 2, alpha, beta, tc0);
00269 }
00270 static void FUNCC(h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
00271 {
00272     FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
00273 }
00274 static void FUNCC(h264_h_loop_filter_chroma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
00275 {
00276     FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 1, alpha, beta, tc0);
00277 }
00278 
00279 static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma_intra)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta)
00280 {
00281     pixel *pix = (pixel*)p_pix;
00282     int d;
00283     xstride >>= sizeof(pixel)-1;
00284     ystride >>= sizeof(pixel)-1;
00285     alpha <<= BIT_DEPTH - 8;
00286     beta  <<= BIT_DEPTH - 8;
00287     for( d = 0; d < 4 * inner_iters; d++ ) {
00288         const int p0 = pix[-1*xstride];
00289         const int p1 = pix[-2*xstride];
00290         const int q0 = pix[0];
00291         const int q1 = pix[1*xstride];
00292 
00293         if( FFABS( p0 - q0 ) < alpha &&
00294             FFABS( p1 - p0 ) < beta &&
00295             FFABS( q1 - q0 ) < beta ) {
00296 
00297             pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
00298             pix[0]        = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
00299         }
00300         pix += ystride;
00301     }
00302 }
00303 static void FUNCC(h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
00304 {
00305     FUNCC(h264_loop_filter_chroma_intra)(pix, stride, sizeof(pixel), 2, alpha, beta);
00306 }
00307 static void FUNCC(h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
00308 {
00309     FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
00310 }
00311 static void FUNCC(h264_h_loop_filter_chroma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
00312 {
00313     FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 1, alpha, beta);
00314 }