falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
29419 Commits
32 Branches
552MB
Tree: a4f6be86d6
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'a4f6be86d6'
${ noResults }
FFmpeg/libavcodec/h264_loopfilter.c

752 lines
37KB
Raw Blame History 

  1. /*

  2.  * H.26L/H.264/AVC/JVT/14496-10/... loop filter

  3.  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>

  4.  *

  5.  * This file is part of Libav.

  6.  *

  7.  * Libav is free software; you can redistribute it and/or

  8.  * modify it under the terms of the GNU Lesser General Public

  9.  * License as published by the Free Software Foundation; either

  10.  * version 2.1 of the License, or (at your option) any later version.

  11.  *

  12.  * Libav is distributed in the hope that it will be useful,

  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  15.  * Lesser General Public License for more details.

  16.  *

  17.  * You should have received a copy of the GNU Lesser General Public

  18.  * License along with Libav; if not, write to the Free Software

  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  20.  */

  21. 

  22. /**

  23.  * @file

  24.  * H.264 / AVC / MPEG4 part10 loop filter.

  25.  * @author Michael Niedermayer <michaelni@gmx.at>

  26.  */

  27. 

  28. #include "libavutil/intreadwrite.h"

  29. #include "internal.h"

  30. #include "dsputil.h"

  31. #include "avcodec.h"

  32. #include "mpegvideo.h"

  33. #include "h264.h"

  34. #include "mathops.h"

  35. #include "rectangle.h"

  36. 

  37. //#undef NDEBUG

  38. #include <assert.h>

  39. 

  40. /* Deblocking filter (p153) */

  41. static const uint8_t alpha_table[52*3] = {

  42.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  43.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  44.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  45.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  46.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  47.      0,  0,  0,  0,  0,  0,  4,  4,  5,  6,

  48.      7,  8,  9, 10, 12, 13, 15, 17, 20, 22,

  49.     25, 28, 32, 36, 40, 45, 50, 56, 63, 71,

  50.     80, 90,101,113,127,144,162,182,203,226,

  51.    255,255,

  52.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  53.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  54.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  55.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  56. };

  57. static const uint8_t beta_table[52*3] = {

  58.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  59.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  60.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  61.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  62.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  63.      0,  0,  0,  0,  0,  0,  2,  2,  2,  3,

  64.      3,  3,  3,  4,  4,  4,  6,  6,  7,  7,

  65.      8,  8,  9,  9, 10, 10, 11, 11, 12, 12,

  66.     13, 13, 14, 14, 15, 15, 16, 16, 17, 17,

  67.     18, 18,

  68.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  69.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  70.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  71.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  72. };

  73. static const uint8_t tc0_table[52*3][4] = {

  74.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  75.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  76.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  77.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  78.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  79.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  80.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  81.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  82.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  83.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  84.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  85.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 1 },

  86.     {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 },

  87.     {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 },

  88.     {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 },

  89.     {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 },

  90.     {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 },

  91.     {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-1,13,17,25 },

  92.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  93.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  94.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  95.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  96.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  97.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  98.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  99.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  100.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  101. };

  102. 

  103. /* intra: 0 if this loopfilter call is guaranteed to be inter (bS < 4), 1 if it might be intra (bS == 4) */

  104. static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) {

  105.     const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  106.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  107.     const int alpha = alpha_table[index_a];

  108.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

  109.     if (alpha ==0 || beta == 0) return;

  110. 

  111.     if( bS[0] < 4 || !intra ) {

  112.         int8_t tc[4];

  113.         tc[0] = tc0_table[index_a][bS[0]];

  114.         tc[1] = tc0_table[index_a][bS[1]];

  115.         tc[2] = tc0_table[index_a][bS[2]];

  116.         tc[3] = tc0_table[index_a][bS[3]];

  117.         h->h264dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);

  118.     } else {

  119.         h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);

  120.     }

  121. }

  122. static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) {

  123.     const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  124.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  125.     const int alpha = alpha_table[index_a];

  126.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

  127.     if (alpha ==0 || beta == 0) return;

  128. 

  129.     if( bS[0] < 4 || !intra ) {

  130.         int8_t tc[4];

  131.         tc[0] = tc0_table[index_a][bS[0]]+1;

  132.         tc[1] = tc0_table[index_a][bS[1]]+1;

  133.         tc[2] = tc0_table[index_a][bS[2]]+1;

  134.         tc[3] = tc0_table[index_a][bS[3]]+1;

  135.         h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);

  136.     } else {

  137.         h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);

  138.     }

  139. }

  140. 

  141. static void av_always_inline filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int intra ) {

  142.     const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  143.     int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  144.     int alpha = alpha_table[index_a];

  145.     int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

  146.     if (alpha ==0 || beta == 0) return;

  147. 

  148.     if( bS[0] < 4 || !intra ) {

  149.         int8_t tc[4];

  150.         tc[0] = tc0_table[index_a][bS[0*bsi]];

  151.         tc[1] = tc0_table[index_a][bS[1*bsi]];

  152.         tc[2] = tc0_table[index_a][bS[2*bsi]];

  153.         tc[3] = tc0_table[index_a][bS[3*bsi]];

  154.         h->h264dsp.h264_h_loop_filter_luma_mbaff(pix, stride, alpha, beta, tc);

  155.     } else {

  156.         h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta);

  157.     }

  158. }

  159. static void av_always_inline filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int intra ) {

  160.     const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  161.     int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  162.     int alpha = alpha_table[index_a];

  163.     int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

  164.     if (alpha ==0 || beta == 0) return;

  165. 

  166.     if( bS[0] < 4 || !intra ) {

  167.         int8_t tc[4];

  168.         tc[0] = tc0_table[index_a][bS[0*bsi]] + 1;

  169.         tc[1] = tc0_table[index_a][bS[1*bsi]] + 1;

  170.         tc[2] = tc0_table[index_a][bS[2*bsi]] + 1;

  171.         tc[3] = tc0_table[index_a][bS[3*bsi]] + 1;

  172.         h->h264dsp.h264_h_loop_filter_chroma_mbaff(pix, stride, alpha, beta, tc);

  173.     } else {

  174.         h->h264dsp.h264_h_loop_filter_chroma_mbaff_intra(pix, stride, alpha, beta);

  175.     }

  176. }

  177. 

  178. static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) {

  179.     const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  180.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  181.     const int alpha = alpha_table[index_a];

  182.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

  183.     if (alpha ==0 || beta == 0) return;

  184. 

  185.     if( bS[0] < 4 || !intra ) {

  186.         int8_t tc[4];

  187.         tc[0] = tc0_table[index_a][bS[0]];

  188.         tc[1] = tc0_table[index_a][bS[1]];

  189.         tc[2] = tc0_table[index_a][bS[2]];

  190.         tc[3] = tc0_table[index_a][bS[3]];

  191.         h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);

  192.     } else {

  193.         h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);

  194.     }

  195. }

  196. 

  197. static void av_always_inline filter_mb_edgech( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) {

  198.     const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  199.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  200.     const int alpha = alpha_table[index_a];

  201.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

  202.     if (alpha ==0 || beta == 0) return;

  203. 

  204.     if( bS[0] < 4 || !intra ) {

  205.         int8_t tc[4];

  206.         tc[0] = tc0_table[index_a][bS[0]]+1;

  207.         tc[1] = tc0_table[index_a][bS[1]]+1;

  208.         tc[2] = tc0_table[index_a][bS[2]]+1;

  209.         tc[3] = tc0_table[index_a][bS[3]]+1;

  210.         h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);

  211.     } else {

  212.         h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);

  213.     }

  214. }

  215. 

  216. void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize) {

  217.     MpegEncContext * const s = &h->s;

  218.     int mb_xy;

  219.     int mb_type, left_type, top_type;

  220.     int qp, qp0, qp1, qpc, qpc0, qpc1;

  221.     int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));

  222.     int chroma444 = CHROMA444;

  223. 

  224.     mb_xy = h->mb_xy;

  225. 

  226.     if(!h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) {

  227.         ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);

  228.         return;

  229.     }

  230.     assert(!FRAME_MBAFF);

  231.     left_type= h->left_type[LTOP];

  232.     top_type= h->top_type;

  233. 

  234.     mb_type = s->current_picture.f.mb_type[mb_xy];

  235.     qp      = s->current_picture.f.qscale_table[mb_xy];

  236.     qp0     = s->current_picture.f.qscale_table[mb_xy - 1];

  237.     qp1     = s->current_picture.f.qscale_table[h->top_mb_xy];

  238.     qpc = get_chroma_qp( h, 0, qp );

  239.     qpc0 = get_chroma_qp( h, 0, qp0 );

  240.     qpc1 = get_chroma_qp( h, 0, qp1 );

  241.     qp0 = (qp + qp0 + 1) >> 1;

  242.     qp1 = (qp + qp1 + 1) >> 1;

  243.     qpc0 = (qpc + qpc0 + 1) >> 1;

  244.     qpc1 = (qpc + qpc1 + 1) >> 1;

  245. 

  246.     if( IS_INTRA(mb_type) ) {

  247.         static const int16_t bS4[4] = {4,4,4,4};

  248.         static const int16_t bS3[4] = {3,3,3,3};

  249.         const int16_t *bSH = FIELD_PICTURE ? bS3 : bS4;

  250.         if(left_type)

  251.             filter_mb_edgev( &img_y[4*0], linesize, bS4, qp0, h, 1);

  252.         if( IS_8x8DCT(mb_type) ) {

  253.             filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, h, 0);

  254.             if(top_type){

  255.                 filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, h, 1);

  256.             }

  257.             filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h, 0);

  258.         } else {

  259.             filter_mb_edgev( &img_y[4*1], linesize, bS3, qp, h, 0);

  260.             filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, h, 0);

  261.             filter_mb_edgev( &img_y[4*3], linesize, bS3, qp, h, 0);

  262.             if(top_type){

  263.                 filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, h, 1);

  264.             }

  265.             filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, h, 0);

  266.             filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h, 0);

  267.             filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, h, 0);

  268.         }

  269.         if(chroma){

  270.             if(chroma444){

  271.                 if(left_type){

  272.                     filter_mb_edgev( &img_cb[4*0], linesize, bS4, qpc0, h, 1);

  273.                     filter_mb_edgev( &img_cr[4*0], linesize, bS4, qpc0, h, 1);

  274.                 }

  275.                 if( IS_8x8DCT(mb_type) ) {

  276.                     filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, h, 0);

  277.                     filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, h, 0);

  278.                     if(top_type){

  279.                         filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, h,1 );

  280.                         filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, h,1 );

  281.                     }

  282.                     filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, h, 0);

  283.                     filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, h, 0);

  284.                 } else {

  285.                     filter_mb_edgev( &img_cb[4*1], linesize, bS3, qpc, h, 0);

  286.                     filter_mb_edgev( &img_cr[4*1], linesize, bS3, qpc, h, 0);

  287.                     filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, h, 0);

  288.                     filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, h, 0);

  289.                     filter_mb_edgev( &img_cb[4*3], linesize, bS3, qpc, h, 0);

  290.                     filter_mb_edgev( &img_cr[4*3], linesize, bS3, qpc, h, 0);

  291.                     if(top_type){

  292.                         filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, h, 1);

  293.                         filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, h, 1);

  294.                     }

  295.                     filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, h, 0);

  296.                     filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, h, 0);

  297.                     filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, h, 0);

  298.                     filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, h, 0);

  299.                     filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, h, 0);

  300.                     filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, h, 0);

  301.                 }

  302.             }else{

  303.                 if(left_type){

  304.                     filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, h, 1);

  305.                     filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, h, 1);

  306.                 }

  307.                 filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, h, 0);

  308.                 filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, h, 0);

  309.                 if(top_type){

  310.                     filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, h, 1);

  311.                     filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, h, 1);

  312.                 }

  313.                 filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, h, 0);

  314.                 filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, h, 0);

  315.             }

  316.         }

  317.         return;

  318.     } else {

  319.         LOCAL_ALIGNED_8(int16_t, bS, [2], [4][4]);

  320.         int edges;

  321.         if( IS_8x8DCT(mb_type) && (h->cbp&7) == 7 && !chroma444 ) {

  322.             edges = 4;

  323.             AV_WN64A(bS[0][0], 0x0002000200020002ULL);

  324.             AV_WN64A(bS[0][2], 0x0002000200020002ULL);

  325.             AV_WN64A(bS[1][0], 0x0002000200020002ULL);

  326.             AV_WN64A(bS[1][2], 0x0002000200020002ULL);

  327.         } else {

  328.             int mask_edge1 = (3*(((5*mb_type)>>5)&1)) | (mb_type>>4); //(mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : (mb_type & MB_TYPE_16x8) ? 1 : 0;

  329.             int mask_edge0 = 3*((mask_edge1>>1) & ((5*left_type)>>5)&1); // (mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) && (h->left_type[LTOP] & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : 0;

  330.             int step =  1+(mb_type>>24); //IS_8x8DCT(mb_type) ? 2 : 1;

  331.             edges = 4 - 3*((mb_type>>3) & !(h->cbp & 15)); //(mb_type & MB_TYPE_16x16) && !(h->cbp & 15) ? 1 : 4;

  332.             h->h264dsp.h264_loop_filter_strength( bS, h->non_zero_count_cache, h->ref_cache, h->mv_cache,

  333.                                               h->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE);

  334.         }

  335.         if( IS_INTRA(left_type) )

  336.             AV_WN64A(bS[0][0], 0x0004000400040004ULL);

  337.         if( IS_INTRA(top_type) )

  338.             AV_WN64A(bS[1][0], FIELD_PICTURE ? 0x0003000300030003ULL : 0x0004000400040004ULL);

  339. 

  340. #define FILTER(hv,dir,edge,intra)\

  341.         if(AV_RN64A(bS[dir][edge])) {                                   \

  342.             filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qp : qp##dir, h, intra );\

  343.             if(chroma){\

  344.                 if(chroma444){\

  345.                     filter_mb_edge##hv( &img_cb[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\

  346.                     filter_mb_edge##hv( &img_cr[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\

  347.                 } else if(!(edge&1)) {\

  348.                     filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\

  349.                     filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\

  350.                 }\

  351.             }\

  352.         }

  353.         if(left_type)

  354.             FILTER(v,0,0,1);

  355.         if( edges == 1 ) {

  356.             if(top_type)

  357.                 FILTER(h,1,0,1);

  358.         } else if( IS_8x8DCT(mb_type) ) {

  359.             FILTER(v,0,2,0);

  360.             if(top_type)

  361.                 FILTER(h,1,0,1);

  362.             FILTER(h,1,2,0);

  363.         } else {

  364.             FILTER(v,0,1,0);

  365.             FILTER(v,0,2,0);

  366.             FILTER(v,0,3,0);

  367.             if(top_type)

  368.                 FILTER(h,1,0,1);

  369.             FILTER(h,1,1,0);

  370.             FILTER(h,1,2,0);

  371.             FILTER(h,1,3,0);

  372.         }

  373. #undef FILTER

  374.     }

  375. }

  376. 

  377. static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){

  378.     int v;

  379. 

  380.     v= h->ref_cache[0][b_idx] != h->ref_cache[0][bn_idx];

  381.     if(!v && h->ref_cache[0][b_idx]!=-1)

  382.         v= h->mv_cache[0][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |

  383.            FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;

  384. 

  385.     if(h->list_count==2){

  386.         if(!v)

  387.             v = h->ref_cache[1][b_idx] != h->ref_cache[1][bn_idx] |

  388.                 h->mv_cache[1][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |

  389.                 FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit;

  390. 

  391.         if(v){

  392.             if(h->ref_cache[0][b_idx] != h->ref_cache[1][bn_idx] |

  393.                h->ref_cache[1][b_idx] != h->ref_cache[0][bn_idx])

  394.                 return 1;

  395.             return

  396.                 h->mv_cache[0][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |

  397.                 FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit |

  398.                 h->mv_cache[1][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |

  399.                 FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;

  400.         }

  401.     }

  402. 

  403.     return v;

  404. }

  405. 

  406. static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int chroma, int chroma444, int dir) {

  407.     MpegEncContext * const s = &h->s;

  408.     int edge;

  409.     int chroma_qp_avg[2];

  410.     const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;

  411.     const int mbm_type = dir == 0 ? h->left_type[LTOP] : h->top_type;

  412. 

  413.     // how often to recheck mv-based bS when iterating between edges

  414.     static const uint8_t mask_edge_tab[2][8]={{0,3,3,3,1,1,1,1},

  415.                                               {0,3,1,1,3,3,3,3}};

  416.     const int mask_edge = mask_edge_tab[dir][(mb_type>>3)&7];

  417.     const int edges = mask_edge== 3 && !(h->cbp&15) ? 1 : 4;

  418. 

  419.     // how often to recheck mv-based bS when iterating along each edge

  420.     const int mask_par0 = mb_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir));

  421. 

  422.     if(mbm_type && !first_vertical_edge_done){

  423. 

  424.         if (FRAME_MBAFF && (dir == 1) && ((mb_y&1) == 0)

  425.             && IS_INTERLACED(mbm_type&~mb_type)

  426.             ) {

  427.             // This is a special case in the norm where the filtering must

  428.             // be done twice (one each of the field) even if we are in a

  429.             // frame macroblock.

  430.             //

  431.             unsigned int tmp_linesize   = 2 *   linesize;

  432.             unsigned int tmp_uvlinesize = 2 * uvlinesize;

  433.             int mbn_xy = mb_xy - 2 * s->mb_stride;

  434.             int j;

  435. 

  436.             for(j=0; j<2; j++, mbn_xy += s->mb_stride){

  437.                 DECLARE_ALIGNED(8, int16_t, bS)[4];

  438.                 int qp;

  439.                 if (IS_INTRA(mb_type | s->current_picture.f.mb_type[mbn_xy])) {

  440.                     AV_WN64A(bS, 0x0003000300030003ULL);

  441.                 } else {

  442.                     if (!CABAC && IS_8x8DCT(s->current_picture.f.mb_type[mbn_xy])) {

  443.                         bS[0]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+0]);

  444.                         bS[1]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+1]);

  445.                         bS[2]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+2]);

  446.                         bS[3]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+3]);

  447.                     }else{

  448.                     const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4;

  449.                     int i;

  450.                     for( i = 0; i < 4; i++ ) {

  451.                         bS[i] = 1 + !!(h->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]);

  452.                     }

  453.                     }

  454.                 }

  455.                 // Do not use s->qscale as luma quantizer because it has not the same

  456.                 // value in IPCM macroblocks.

  457.                 qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbn_xy] + 1) >> 1;

  458.                 tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);

  459.                 { int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  460.                 filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, h, 0 );

  461.                 chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;

  462.                 chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;

  463.                 if (chroma) {

  464.                     if (chroma444) {

  465.                         filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h, 0);

  466.                         filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h, 0);

  467.                     } else {

  468.                         filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h, 0);

  469.                         filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h, 0);

  470.                     }

  471.                 }

  472.             }

  473.         }else{

  474.             DECLARE_ALIGNED(8, int16_t, bS)[4];

  475.             int qp;

  476. 

  477.             if( IS_INTRA(mb_type|mbm_type)) {

  478.                 AV_WN64A(bS, 0x0003000300030003ULL);

  479.                 if (   (!IS_INTERLACED(mb_type|mbm_type))

  480.                     || ((FRAME_MBAFF || (s->picture_structure != PICT_FRAME)) && (dir == 0))

  481.                 )

  482.                     AV_WN64A(bS, 0x0004000400040004ULL);

  483.             } else {

  484.                 int i;

  485.                 int mv_done;

  486. 

  487.                 if( dir && FRAME_MBAFF && IS_INTERLACED(mb_type ^ mbm_type)) {

  488.                     AV_WN64A(bS, 0x0001000100010001ULL);

  489.                     mv_done = 1;

  490.                 }

  491.                 else if( mask_par0 && ((mbm_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) {

  492.                     int b_idx= 8 + 4;

  493.                     int bn_idx= b_idx - (dir ? 8:1);

  494. 

  495.                     bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, 8 + 4, bn_idx, mvy_limit);

  496.                     mv_done = 1;

  497.                 }

  498.                 else

  499.                     mv_done = 0;

  500. 

  501.                 for( i = 0; i < 4; i++ ) {

  502.                     int x = dir == 0 ? 0 : i;

  503.                     int y = dir == 0 ? i    : 0;

  504.                     int b_idx= 8 + 4 + x + 8*y;

  505.                     int bn_idx= b_idx - (dir ? 8:1);

  506. 

  507.                     if( h->non_zero_count_cache[b_idx] |

  508.                         h->non_zero_count_cache[bn_idx] ) {

  509.                         bS[i] = 2;

  510.                     }

  511.                     else if(!mv_done)

  512.                     {

  513.                         bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);

  514.                     }

  515.                 }

  516.             }

  517. 

  518.             /* Filter edge */

  519.             // Do not use s->qscale as luma quantizer because it has not the same

  520.             // value in IPCM macroblocks.

  521.             if(bS[0]+bS[1]+bS[2]+bS[3]){

  522.                 qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbm_xy] + 1) >> 1;

  523.                 //tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]);

  524.                 tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);

  525.                 //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  526.                 chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;

  527.                 chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;

  528.                 if( dir == 0 ) {

  529.                     filter_mb_edgev( &img_y[0], linesize, bS, qp, h, 1 );

  530.                     if (chroma) {

  531.                         if (chroma444) {

  532.                             filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1);

  533.                             filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1);

  534.                         } else {

  535.                             filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1);

  536.                             filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1);

  537.                         }

  538.                     }

  539.                 } else {

  540.                     filter_mb_edgeh( &img_y[0], linesize, bS, qp, h, 1 );

  541.                     if (chroma) {

  542.                         if (chroma444) {

  543.                             filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1);

  544.                             filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1);

  545.                         } else {

  546.                             filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1);

  547.                             filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1);

  548.                         }

  549.                     }

  550.                 }

  551.             }

  552.         }

  553.     }

  554. 

  555.     /* Calculate bS */

  556.     for( edge = 1; edge < edges; edge++ ) {

  557.         DECLARE_ALIGNED(8, int16_t, bS)[4];

  558.         int qp;

  559. 

  560.         if( IS_8x8DCT(mb_type & (edge<<24)) ) // (edge&1) && IS_8x8DCT(mb_type)

  561.             continue;

  562. 

  563.         if( IS_INTRA(mb_type)) {

  564.             AV_WN64A(bS, 0x0003000300030003ULL);

  565.         } else {

  566.             int i;

  567.             int mv_done;

  568. 

  569.             if( edge & mask_edge ) {

  570.                 AV_ZERO64(bS);

  571.                 mv_done = 1;

  572.             }

  573.             else if( mask_par0 ) {

  574.                 int b_idx= 8 + 4 + edge * (dir ? 8:1);

  575.                 int bn_idx= b_idx - (dir ? 8:1);

  576. 

  577.                 bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, b_idx, bn_idx, mvy_limit);

  578.                 mv_done = 1;

  579.             }

  580.             else

  581.                 mv_done = 0;

  582. 

  583.             for( i = 0; i < 4; i++ ) {

  584.                 int x = dir == 0 ? edge : i;

  585.                 int y = dir == 0 ? i    : edge;

  586.                 int b_idx= 8 + 4 + x + 8*y;

  587.                 int bn_idx= b_idx - (dir ? 8:1);

  588. 

  589.                 if( h->non_zero_count_cache[b_idx] |

  590.                     h->non_zero_count_cache[bn_idx] ) {

  591.                     bS[i] = 2;

  592.                 }

  593.                 else if(!mv_done)

  594.                 {

  595.                     bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);

  596.                 }

  597.             }

  598. 

  599.             if(bS[0]+bS[1]+bS[2]+bS[3] == 0)

  600.                 continue;

  601.         }

  602. 

  603.         /* Filter edge */

  604.         // Do not use s->qscale as luma quantizer because it has not the same

  605.         // value in IPCM macroblocks.

  606.         qp = s->current_picture.f.qscale_table[mb_xy];

  607.         //tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]);

  608.         tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);

  609.         //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  610.         if( dir == 0 ) {

  611.             filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, h, 0 );

  612.             if (chroma) {

  613.                 if (chroma444) {

  614.                     filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h, 0);

  615.                     filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h, 0);

  616.                 } else if( (edge&1) == 0 ) {

  617.                     filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h, 0);

  618.                     filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h, 0);

  619.                 }

  620.             }

  621.         } else {

  622.             filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, h, 0 );

  623.             if (chroma) {

  624.                 if (chroma444) {

  625.                     filter_mb_edgeh ( &img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h, 0);

  626.                     filter_mb_edgeh ( &img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h, 0);

  627.                 } else if( (edge&1) == 0 ) {

  628.                     filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h, 0);

  629.                     filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h, 0);

  630.                 }

  631.             }

  632.         }

  633.     }

  634. }

  635. 

  636. void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize) {

  637.     MpegEncContext * const s = &h->s;

  638.     const int mb_xy= mb_x + mb_y*s->mb_stride;

  639.     const int mb_type = s->current_picture.f.mb_type[mb_xy];

  640.     const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4;

  641.     int first_vertical_edge_done = 0;

  642.     av_unused int dir;

  643.     int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));

  644. 

  645.     if (FRAME_MBAFF

  646.             // and current and left pair do not have the same interlaced type

  647.             && IS_INTERLACED(mb_type^h->left_type[LTOP])

  648.             // and left mb is in available to us

  649.             && h->left_type[LTOP]) {

  650.         /* First vertical edge is different in MBAFF frames

  651.          * There are 8 different bS to compute and 2 different Qp

  652.          */

  653.         DECLARE_ALIGNED(8, int16_t, bS)[8];

  654.         int qp[2];

  655.         int bqp[2];

  656.         int rqp[2];

  657.         int mb_qp, mbn0_qp, mbn1_qp;

  658.         int i;

  659.         first_vertical_edge_done = 1;

  660. 

  661.         if( IS_INTRA(mb_type) ) {

  662.             AV_WN64A(&bS[0], 0x0004000400040004ULL);

  663.             AV_WN64A(&bS[4], 0x0004000400040004ULL);

  664.         } else {

  665.             static const uint8_t offset[2][2][8]={

  666.                 {

  667.                     {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},

  668.                     {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},

  669.                 },{

  670.                     {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},

  671.                     {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},

  672.                 }

  673.             };

  674.             const uint8_t *off= offset[MB_FIELD][mb_y&1];

  675.             for( i = 0; i < 8; i++ ) {

  676.                 int j= MB_FIELD ? i>>2 : i&1;

  677.                 int mbn_xy = h->left_mb_xy[LEFT(j)];

  678.                 int mbn_type= h->left_type[LEFT(j)];

  679. 

  680.                 if( IS_INTRA( mbn_type ) )

  681.                     bS[i] = 4;

  682.                 else{

  683.                     bS[i] = 1 + !!(h->non_zero_count_cache[12+8*(i>>1)] |

  684.                          ((!h->pps.cabac && IS_8x8DCT(mbn_type)) ?

  685.                             (h->cbp_table[mbn_xy] & (((MB_FIELD ? (i&2) : (mb_y&1)) ? 8 : 2) << 12))

  686.                                                                        :

  687.                             h->non_zero_count[mbn_xy][ off[i] ]));

  688.                 }

  689.             }

  690.         }

  691. 

  692.         mb_qp   = s->current_picture.f.qscale_table[mb_xy];

  693.         mbn0_qp = s->current_picture.f.qscale_table[h->left_mb_xy[0]];

  694.         mbn1_qp = s->current_picture.f.qscale_table[h->left_mb_xy[1]];

  695.         qp[0] = ( mb_qp + mbn0_qp + 1 ) >> 1;

  696.         bqp[0] = ( get_chroma_qp( h, 0, mb_qp ) +

  697.                    get_chroma_qp( h, 0, mbn0_qp ) + 1 ) >> 1;

  698.         rqp[0] = ( get_chroma_qp( h, 1, mb_qp ) +

  699.                    get_chroma_qp( h, 1, mbn0_qp ) + 1 ) >> 1;

  700.         qp[1] = ( mb_qp + mbn1_qp + 1 ) >> 1;

  701.         bqp[1] = ( get_chroma_qp( h, 0, mb_qp ) +

  702.                    get_chroma_qp( h, 0, mbn1_qp ) + 1 ) >> 1;

  703.         rqp[1] = ( get_chroma_qp( h, 1, mb_qp ) +

  704.                    get_chroma_qp( h, 1, mbn1_qp ) + 1 ) >> 1;

  705. 

  706.         /* Filter edge */

  707.         tprintf(s->avctx, "filter mb:%d/%d MBAFF, QPy:%d/%d, QPb:%d/%d QPr:%d/%d ls:%d uvls:%d", mb_x, mb_y, qp[0], qp[1], bqp[0], bqp[1], rqp[0], rqp[1], linesize, uvlinesize);

  708.         { int i; for (i = 0; i < 8; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  709.         if(MB_FIELD){

  710.             filter_mb_mbaff_edgev ( h, img_y                ,   linesize, bS  , 1, qp [0], 1 );

  711.             filter_mb_mbaff_edgev ( h, img_y  + 8*  linesize,   linesize, bS+4, 1, qp [1], 1 );

  712.             if (chroma){

  713.                 if (CHROMA444) {

  714.                     filter_mb_mbaff_edgev ( h, img_cb,                uvlinesize, bS  , 1, bqp[0], 1 );

  715.                     filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], 1 );

  716.                     filter_mb_mbaff_edgev ( h, img_cr,                uvlinesize, bS  , 1, rqp[0], 1 );

  717.                     filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], 1 );

  718.                 }else{

  719.                     filter_mb_mbaff_edgecv( h, img_cb,                uvlinesize, bS  , 1, bqp[0], 1 );

  720.                     filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], 1 );

  721.                     filter_mb_mbaff_edgecv( h, img_cr,                uvlinesize, bS  , 1, rqp[0], 1 );

  722.                     filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], 1 );

  723.                 }

  724.             }

  725.         }else{

  726.             filter_mb_mbaff_edgev ( h, img_y              , 2*  linesize, bS  , 2, qp [0], 1 );

  727.             filter_mb_mbaff_edgev ( h, img_y  +   linesize, 2*  linesize, bS+1, 2, qp [1], 1 );

  728.             if (chroma){

  729.                 if (CHROMA444) {

  730.                     filter_mb_mbaff_edgev ( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0], 1 );

  731.                     filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], 1 );

  732.                     filter_mb_mbaff_edgev ( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0], 1 );

  733.                     filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], 1 );

  734.                 }else{

  735.                     filter_mb_mbaff_edgecv( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0], 1 );

  736.                     filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], 1 );

  737.                     filter_mb_mbaff_edgecv( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0], 1 );

  738.                     filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], 1 );

  739.                 }

  740.             }

  741.         }

  742.     }

  743. 

  744. #if CONFIG_SMALL

  745.     for( dir = 0; dir < 2; dir++ )

  746.         filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, chroma, CHROMA444, dir);

  747. #else

  748.     filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, chroma, CHROMA444, 0);

  749.     filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, chroma, CHROMA444, 1);

  750. #endif

  751. }