h264_loopfilter.c
Go to the documentation of this file.
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 
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) */
105  const int16_t bS[4],
106  unsigned int qp, int a, int b,
107  H264Context *h, int intra)
108 {
109  const unsigned int index_a = qp + a;
110  const int alpha = alpha_table[index_a];
111  const int beta = beta_table[qp + b];
112  if (alpha ==0 || beta == 0) return;
113 
114  if( bS[0] < 4 || !intra ) {
115  int8_t tc[4];
116  tc[0] = tc0_table[index_a][bS[0]];
117  tc[1] = tc0_table[index_a][bS[1]];
118  tc[2] = tc0_table[index_a][bS[2]];
119  tc[3] = tc0_table[index_a][bS[3]];
120  h->h264dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);
121  } else {
122  h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);
123  }
124 }
125 
127  const int16_t bS[4],
128  unsigned int qp, int a, int b,
129  H264Context *h, int intra)
130 {
131  const unsigned int index_a = qp + a;
132  const int alpha = alpha_table[index_a];
133  const int beta = beta_table[qp + b];
134  if (alpha ==0 || beta == 0) return;
135 
136  if( bS[0] < 4 || !intra ) {
137  int8_t tc[4];
138  tc[0] = tc0_table[index_a][bS[0]]+1;
139  tc[1] = tc0_table[index_a][bS[1]]+1;
140  tc[2] = tc0_table[index_a][bS[2]]+1;
141  tc[3] = tc0_table[index_a][bS[3]]+1;
142  h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);
143  } else {
144  h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);
145  }
146 }
147 
149  int stride,
150  const int16_t bS[7], int bsi,
151  int qp, int a, int b,
152  int intra)
153 {
154  const unsigned int index_a = qp + a;
155  const int alpha = alpha_table[index_a];
156  const int beta = beta_table[qp + b];
157  if (alpha ==0 || beta == 0) return;
158 
159  if( bS[0] < 4 || !intra ) {
160  int8_t tc[4];
161  tc[0] = tc0_table[index_a][bS[0*bsi]];
162  tc[1] = tc0_table[index_a][bS[1*bsi]];
163  tc[2] = tc0_table[index_a][bS[2*bsi]];
164  tc[3] = tc0_table[index_a][bS[3*bsi]];
165  h->h264dsp.h264_h_loop_filter_luma_mbaff(pix, stride, alpha, beta, tc);
166  } else {
167  h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta);
168  }
169 }
170 
172  uint8_t *pix, int stride,
173  const int16_t bS[7],
174  int bsi, int qp, int a,
175  int b, int intra)
176 {
177  const unsigned int index_a = qp + a;
178  const int alpha = alpha_table[index_a];
179  const int beta = beta_table[qp + b];
180  if (alpha ==0 || beta == 0) return;
181 
182  if( bS[0] < 4 || !intra ) {
183  int8_t tc[4];
184  tc[0] = tc0_table[index_a][bS[0*bsi]] + 1;
185  tc[1] = tc0_table[index_a][bS[1*bsi]] + 1;
186  tc[2] = tc0_table[index_a][bS[2*bsi]] + 1;
187  tc[3] = tc0_table[index_a][bS[3*bsi]] + 1;
188  h->h264dsp.h264_h_loop_filter_chroma_mbaff(pix, stride, alpha, beta, tc);
189  } else {
190  h->h264dsp.h264_h_loop_filter_chroma_mbaff_intra(pix, stride, alpha, beta);
191  }
192 }
193 
195  const int16_t bS[4],
196  unsigned int qp, int a, int b,
197  H264Context *h, int intra)
198 {
199  const unsigned int index_a = qp + a;
200  const int alpha = alpha_table[index_a];
201  const int beta = beta_table[qp + b];
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]];
207  tc[1] = tc0_table[index_a][bS[1]];
208  tc[2] = tc0_table[index_a][bS[2]];
209  tc[3] = tc0_table[index_a][bS[3]];
210  h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
211  } else {
212  h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);
213  }
214 }
215 
217  const int16_t bS[4],
218  unsigned int qp, int a, int b,
219  H264Context *h, int intra)
220 {
221  const unsigned int index_a = qp + a;
222  const int alpha = alpha_table[index_a];
223  const int beta = beta_table[qp + b];
224  if (alpha ==0 || beta == 0) return;
225 
226  if( bS[0] < 4 || !intra ) {
227  int8_t tc[4];
228  tc[0] = tc0_table[index_a][bS[0]]+1;
229  tc[1] = tc0_table[index_a][bS[1]]+1;
230  tc[2] = tc0_table[index_a][bS[2]]+1;
231  tc[3] = tc0_table[index_a][bS[3]]+1;
232  h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);
233  } else {
234  h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);
235  }
236 }
237 
239  int mb_x, int mb_y,
240  uint8_t *img_y,
241  uint8_t *img_cb,
242  uint8_t *img_cr,
243  unsigned int linesize,
244  unsigned int uvlinesize,
245  int pixel_shift)
246 {
247  int chroma = !(CONFIG_GRAY && (h->flags&CODEC_FLAG_GRAY));
248  int chroma444 = CHROMA444;
249  int chroma422 = CHROMA422;
250 
251  int mb_xy = h->mb_xy;
252  int left_type= h->left_type[LTOP];
253  int top_type= h->top_type;
254 
255  int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
256  int a = 52 + h->slice_alpha_c0_offset - qp_bd_offset;
257  int b = 52 + h->slice_beta_offset - qp_bd_offset;
258 
259  int mb_type = h->cur_pic.f.mb_type[mb_xy];
260  int qp = h->cur_pic.f.qscale_table[mb_xy];
261  int qp0 = h->cur_pic.f.qscale_table[mb_xy - 1];
262  int qp1 = h->cur_pic.f.qscale_table[h->top_mb_xy];
263  int qpc = get_chroma_qp( h, 0, qp );
264  int qpc0 = get_chroma_qp( h, 0, qp0 );
265  int qpc1 = get_chroma_qp( h, 0, qp1 );
266  qp0 = (qp + qp0 + 1) >> 1;
267  qp1 = (qp + qp1 + 1) >> 1;
268  qpc0 = (qpc + qpc0 + 1) >> 1;
269  qpc1 = (qpc + qpc1 + 1) >> 1;
270 
271  if( IS_INTRA(mb_type) ) {
272  static const int16_t bS4[4] = {4,4,4,4};
273  static const int16_t bS3[4] = {3,3,3,3};
274  const int16_t *bSH = FIELD_PICTURE ? bS3 : bS4;
275  if(left_type)
276  filter_mb_edgev( &img_y[4*0<<pixel_shift], linesize, bS4, qp0, a, b, h, 1);
277  if( IS_8x8DCT(mb_type) ) {
278  filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
279  if(top_type){
280  filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
281  }
282  filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
283  } else {
284  filter_mb_edgev( &img_y[4*1<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
285  filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
286  filter_mb_edgev( &img_y[4*3<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
287  if(top_type){
288  filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
289  }
290  filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, a, b, h, 0);
291  filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
292  filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, a, b, h, 0);
293  }
294  if(chroma){
295  if(chroma444){
296  if(left_type){
297  filter_mb_edgev( &img_cb[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1);
298  filter_mb_edgev( &img_cr[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1);
299  }
300  if( IS_8x8DCT(mb_type) ) {
301  filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
302  filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
303  if(top_type){
304  filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
305  filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
306  }
307  filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
308  filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
309  } else {
310  filter_mb_edgev( &img_cb[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
311  filter_mb_edgev( &img_cr[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
312  filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
313  filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
314  filter_mb_edgev( &img_cb[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
315  filter_mb_edgev( &img_cr[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
316  if(top_type){
317  filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
318  filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
319  }
320  filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
321  filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
322  filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
323  filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
324  filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
325  filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
326  }
327  }else if(chroma422){
328  if(left_type){
329  filter_mb_edgecv(&img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
330  filter_mb_edgecv(&img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
331  }
332  filter_mb_edgecv(&img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
333  filter_mb_edgecv(&img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
334  if(top_type){
335  filter_mb_edgech(&img_cb[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
336  filter_mb_edgech(&img_cr[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
337  }
338  filter_mb_edgech(&img_cb[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
339  filter_mb_edgech(&img_cr[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
340  filter_mb_edgech(&img_cb[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
341  filter_mb_edgech(&img_cr[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
342  filter_mb_edgech(&img_cb[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
343  filter_mb_edgech(&img_cr[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
344  }else{
345  if(left_type){
346  filter_mb_edgecv( &img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
347  filter_mb_edgecv( &img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
348  }
349  filter_mb_edgecv( &img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
350  filter_mb_edgecv( &img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
351  if(top_type){
352  filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
353  filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
354  }
355  filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
356  filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
357  }
358  }
359  return;
360  } else {
361  LOCAL_ALIGNED_8(int16_t, bS, [2], [4][4]);
362  int edges;
363  if( IS_8x8DCT(mb_type) && (h->cbp&7) == 7 && !chroma444 ) {
364  edges = 4;
365  AV_WN64A(bS[0][0], 0x0002000200020002ULL);
366  AV_WN64A(bS[0][2], 0x0002000200020002ULL);
367  AV_WN64A(bS[1][0], 0x0002000200020002ULL);
368  AV_WN64A(bS[1][2], 0x0002000200020002ULL);
369  } else {
370  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;
371  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;
372  int step = 1+(mb_type>>24); //IS_8x8DCT(mb_type) ? 2 : 1;
373  edges = 4 - 3*((mb_type>>3) & !(h->cbp & 15)); //(mb_type & MB_TYPE_16x16) && !(h->cbp & 15) ? 1 : 4;
375  h->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE);
376  }
377  if( IS_INTRA(left_type) )
378  AV_WN64A(bS[0][0], 0x0004000400040004ULL);
379  if( IS_INTRA(top_type) )
380  AV_WN64A(bS[1][0], FIELD_PICTURE ? 0x0003000300030003ULL : 0x0004000400040004ULL);
381 
382 #define FILTER(hv,dir,edge,intra)\
383  if(AV_RN64A(bS[dir][edge])) { \
384  filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1<<pixel_shift)], linesize, bS[dir][edge], edge ? qp : qp##dir, a, b, h, intra );\
385  if(chroma){\
386  if(chroma444){\
387  filter_mb_edge##hv( &img_cb[4*edge*(dir?linesize:1<<pixel_shift)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
388  filter_mb_edge##hv( &img_cr[4*edge*(dir?linesize:1<<pixel_shift)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
389  } else if(!(edge&1)) {\
390  filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1<<pixel_shift)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
391  filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1<<pixel_shift)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
392  }\
393  }\
394  }
395  if(left_type)
396  FILTER(v,0,0,1);
397  if( edges == 1 ) {
398  if(top_type)
399  FILTER(h,1,0,1);
400  } else if( IS_8x8DCT(mb_type) ) {
401  FILTER(v,0,2,0);
402  if(top_type)
403  FILTER(h,1,0,1);
404  FILTER(h,1,2,0);
405  } else {
406  FILTER(v,0,1,0);
407  FILTER(v,0,2,0);
408  FILTER(v,0,3,0);
409  if(top_type)
410  FILTER(h,1,0,1);
411  FILTER(h,1,1,0);
412  FILTER(h,1,2,0);
413  FILTER(h,1,3,0);
414  }
415 #undef FILTER
416  }
417 }
418 
419 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) {
420  assert(!FRAME_MBAFF);
422  ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);
423  return;
424  }
425 
426 #if CONFIG_SMALL
427  h264_filter_mb_fast_internal(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, h->pixel_shift);
428 #else
429  if(h->pixel_shift){
430  h264_filter_mb_fast_internal(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 1);
431  }else{
432  h264_filter_mb_fast_internal(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 0);
433  }
434 #endif
435 }
436 
437 static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){
438  int v;
439 
440  v= h->ref_cache[0][b_idx] != h->ref_cache[0][bn_idx];
441  if(!v && h->ref_cache[0][b_idx]!=-1)
442  v= h->mv_cache[0][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |
443  FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;
444 
445  if(h->list_count==2){
446  if(!v)
447  v = h->ref_cache[1][b_idx] != h->ref_cache[1][bn_idx] |
448  h->mv_cache[1][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |
449  FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit;
450 
451  if(v){
452  if(h->ref_cache[0][b_idx] != h->ref_cache[1][bn_idx] |
453  h->ref_cache[1][b_idx] != h->ref_cache[0][bn_idx])
454  return 1;
455  return
456  h->mv_cache[0][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |
457  FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit |
458  h->mv_cache[1][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |
459  FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;
460  }
461  }
462 
463  return v;
464 }
465 
466 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 a, int b, int chroma, int dir) {
467  int edge;
468  int chroma_qp_avg[2];
469  int chroma444 = CHROMA444;
470  int chroma422 = CHROMA422;
471  const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;
472  const int mbm_type = dir == 0 ? h->left_type[LTOP] : h->top_type;
473 
474  // how often to recheck mv-based bS when iterating between edges
475  static const uint8_t mask_edge_tab[2][8]={{0,3,3,3,1,1,1,1},
476  {0,3,1,1,3,3,3,3}};
477  const int mask_edge = mask_edge_tab[dir][(mb_type>>3)&7];
478  const int edges = mask_edge== 3 && !(h->cbp&15) ? 1 : 4;
479 
480  // how often to recheck mv-based bS when iterating along each edge
481  const int mask_par0 = mb_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir));
482 
483  if(mbm_type && !first_vertical_edge_done){
484 
485  if (FRAME_MBAFF && (dir == 1) && ((mb_y&1) == 0)
486  && IS_INTERLACED(mbm_type&~mb_type)
487  ) {
488  // This is a special case in the norm where the filtering must
489  // be done twice (one each of the field) even if we are in a
490  // frame macroblock.
491  //
492  unsigned int tmp_linesize = 2 * linesize;
493  unsigned int tmp_uvlinesize = 2 * uvlinesize;
494  int mbn_xy = mb_xy - 2 * h->mb_stride;
495  int j;
496 
497  for(j=0; j<2; j++, mbn_xy += h->mb_stride){
498  DECLARE_ALIGNED(8, int16_t, bS)[4];
499  int qp;
500  if (IS_INTRA(mb_type | h->cur_pic.f.mb_type[mbn_xy])) {
501  AV_WN64A(bS, 0x0003000300030003ULL);
502  } else {
503  if (!CABAC && IS_8x8DCT(h->cur_pic.f.mb_type[mbn_xy])) {
504  bS[0]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+0]);
505  bS[1]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+1]);
506  bS[2]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+2]);
507  bS[3]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+3]);
508  }else{
509  const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4;
510  int i;
511  for( i = 0; i < 4; i++ ) {
512  bS[i] = 1 + !!(h->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]);
513  }
514  }
515  }
516  // Do not use s->qscale as luma quantizer because it has not the same
517  // value in IPCM macroblocks.
518  qp = (h->cur_pic.f.qscale_table[mb_xy] + h->cur_pic.f.qscale_table[mbn_xy] + 1) >> 1;
519  tprintf(h->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);
520  { int i; for (i = 0; i < 4; i++) tprintf(h->avctx, " bS[%d]:%d", i, bS[i]); tprintf(h->avctx, "\n"); }
521  filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 );
522  chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.f.qscale_table[mbn_xy]) + 1) >> 1;
523  chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.f.qscale_table[mbn_xy]) + 1) >> 1;
524  if (chroma) {
525  if (chroma444) {
526  filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
527  filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
528  } else {
529  filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
530  filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
531  }
532  }
533  }
534  }else{
535  DECLARE_ALIGNED(8, int16_t, bS)[4];
536  int qp;
537 
538  if( IS_INTRA(mb_type|mbm_type)) {
539  AV_WN64A(bS, 0x0003000300030003ULL);
540  if ( (!IS_INTERLACED(mb_type|mbm_type))
541  || ((FRAME_MBAFF || (h->picture_structure != PICT_FRAME)) && (dir == 0))
542  )
543  AV_WN64A(bS, 0x0004000400040004ULL);
544  } else {
545  int i;
546  int mv_done;
547 
548  if( dir && FRAME_MBAFF && IS_INTERLACED(mb_type ^ mbm_type)) {
549  AV_WN64A(bS, 0x0001000100010001ULL);
550  mv_done = 1;
551  }
552  else if( mask_par0 && ((mbm_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) {
553  int b_idx= 8 + 4;
554  int bn_idx= b_idx - (dir ? 8:1);
555 
556  bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, 8 + 4, bn_idx, mvy_limit);
557  mv_done = 1;
558  }
559  else
560  mv_done = 0;
561 
562  for( i = 0; i < 4; i++ ) {
563  int x = dir == 0 ? 0 : i;
564  int y = dir == 0 ? i : 0;
565  int b_idx= 8 + 4 + x + 8*y;
566  int bn_idx= b_idx - (dir ? 8:1);
567 
568  if( h->non_zero_count_cache[b_idx] |
569  h->non_zero_count_cache[bn_idx] ) {
570  bS[i] = 2;
571  }
572  else if(!mv_done)
573  {
574  bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);
575  }
576  }
577  }
578 
579  /* Filter edge */
580  // Do not use s->qscale as luma quantizer because it has not the same
581  // value in IPCM macroblocks.
582  if(bS[0]+bS[1]+bS[2]+bS[3]){
583  qp = (h->cur_pic.f.qscale_table[mb_xy] + h->cur_pic.f.qscale_table[mbm_xy] + 1) >> 1;
584  tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
585  chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.f.qscale_table[mbm_xy]) + 1) >> 1;
586  chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.f.qscale_table[mbm_xy]) + 1) >> 1;
587  if( dir == 0 ) {
588  filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 );
589  if (chroma) {
590  if (chroma444) {
591  filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
592  filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
593  } else {
594  filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
595  filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
596  }
597  }
598  } else {
599  filter_mb_edgeh( &img_y[0], linesize, bS, qp, a, b, h, 1 );
600  if (chroma) {
601  if (chroma444) {
602  filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
603  filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
604  } else {
605  filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
606  filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
607  }
608  }
609  }
610  }
611  }
612  }
613 
614  /* Calculate bS */
615  for( edge = 1; edge < edges; edge++ ) {
616  DECLARE_ALIGNED(8, int16_t, bS)[4];
617  int qp;
618  const int deblock_edge = !IS_8x8DCT(mb_type & (edge<<24)); // (edge&1) && IS_8x8DCT(mb_type)
619 
620  if (!deblock_edge && (!chroma422 || dir == 0))
621  continue;
622 
623  if( IS_INTRA(mb_type)) {
624  AV_WN64A(bS, 0x0003000300030003ULL);
625  } else {
626  int i;
627  int mv_done;
628 
629  if( edge & mask_edge ) {
630  AV_ZERO64(bS);
631  mv_done = 1;
632  }
633  else if( mask_par0 ) {
634  int b_idx= 8 + 4 + edge * (dir ? 8:1);
635  int bn_idx= b_idx - (dir ? 8:1);
636 
637  bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, b_idx, bn_idx, mvy_limit);
638  mv_done = 1;
639  }
640  else
641  mv_done = 0;
642 
643  for( i = 0; i < 4; i++ ) {
644  int x = dir == 0 ? edge : i;
645  int y = dir == 0 ? i : edge;
646  int b_idx= 8 + 4 + x + 8*y;
647  int bn_idx= b_idx - (dir ? 8:1);
648 
649  if( h->non_zero_count_cache[b_idx] |
650  h->non_zero_count_cache[bn_idx] ) {
651  bS[i] = 2;
652  }
653  else if(!mv_done)
654  {
655  bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);
656  }
657  }
658 
659  if(bS[0]+bS[1]+bS[2]+bS[3] == 0)
660  continue;
661  }
662 
663  /* Filter edge */
664  // Do not use s->qscale as luma quantizer because it has not the same
665  // value in IPCM macroblocks.
666  qp = h->cur_pic.f.qscale_table[mb_xy];
667  tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
668  if( dir == 0 ) {
669  filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 );
670  if (chroma) {
671  if (chroma444) {
672  filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
673  filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
674  } else if( (edge&1) == 0 ) {
675  filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
676  filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
677  }
678  }
679  } else {
680  if (chroma422) {
681  if (deblock_edge)
682  filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
683  if (chroma) {
684  filter_mb_edgech(&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
685  filter_mb_edgech(&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
686  }
687  } else {
688  filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
689  if (chroma) {
690  if (chroma444) {
691  filter_mb_edgeh (&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
692  filter_mb_edgeh (&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
693  } else if ((edge&1) == 0) {
694  filter_mb_edgech(&img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
695  filter_mb_edgech(&img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
696  }
697  }
698  }
699  }
700  }
701 }
702 
703 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) {
704  const int mb_xy= mb_x + mb_y*h->mb_stride;
705  const int mb_type = h->cur_pic.f.mb_type[mb_xy];
706  const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4;
707  int first_vertical_edge_done = 0;
708  av_unused int dir;
709  int chroma = !(CONFIG_GRAY && (h->flags&CODEC_FLAG_GRAY));
710  int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
711  int a = 52 + h->slice_alpha_c0_offset - qp_bd_offset;
712  int b = 52 + h->slice_beta_offset - qp_bd_offset;
713 
714  if (FRAME_MBAFF
715  // and current and left pair do not have the same interlaced type
716  && IS_INTERLACED(mb_type^h->left_type[LTOP])
717  // and left mb is in available to us
718  && h->left_type[LTOP]) {
719  /* First vertical edge is different in MBAFF frames
720  * There are 8 different bS to compute and 2 different Qp
721  */
722  DECLARE_ALIGNED(8, int16_t, bS)[8];
723  int qp[2];
724  int bqp[2];
725  int rqp[2];
726  int mb_qp, mbn0_qp, mbn1_qp;
727  int i;
728  first_vertical_edge_done = 1;
729 
730  if( IS_INTRA(mb_type) ) {
731  AV_WN64A(&bS[0], 0x0004000400040004ULL);
732  AV_WN64A(&bS[4], 0x0004000400040004ULL);
733  } else {
734  static const uint8_t offset[2][2][8]={
735  {
736  {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},
737  {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},
738  },{
739  {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
740  {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
741  }
742  };
743  const uint8_t *off= offset[MB_FIELD][mb_y&1];
744  for( i = 0; i < 8; i++ ) {
745  int j= MB_FIELD ? i>>2 : i&1;
746  int mbn_xy = h->left_mb_xy[LEFT(j)];
747  int mbn_type= h->left_type[LEFT(j)];
748 
749  if( IS_INTRA( mbn_type ) )
750  bS[i] = 4;
751  else{
752  bS[i] = 1 + !!(h->non_zero_count_cache[12+8*(i>>1)] |
753  ((!h->pps.cabac && IS_8x8DCT(mbn_type)) ?
754  (h->cbp_table[mbn_xy] & (((MB_FIELD ? (i&2) : (mb_y&1)) ? 8 : 2) << 12))
755  :
756  h->non_zero_count[mbn_xy][ off[i] ]));
757  }
758  }
759  }
760 
761  mb_qp = h->cur_pic.f.qscale_table[mb_xy];
762  mbn0_qp = h->cur_pic.f.qscale_table[h->left_mb_xy[0]];
763  mbn1_qp = h->cur_pic.f.qscale_table[h->left_mb_xy[1]];
764  qp[0] = ( mb_qp + mbn0_qp + 1 ) >> 1;
765  bqp[0] = ( get_chroma_qp( h, 0, mb_qp ) +
766  get_chroma_qp( h, 0, mbn0_qp ) + 1 ) >> 1;
767  rqp[0] = ( get_chroma_qp( h, 1, mb_qp ) +
768  get_chroma_qp( h, 1, mbn0_qp ) + 1 ) >> 1;
769  qp[1] = ( mb_qp + mbn1_qp + 1 ) >> 1;
770  bqp[1] = ( get_chroma_qp( h, 0, mb_qp ) +
771  get_chroma_qp( h, 0, mbn1_qp ) + 1 ) >> 1;
772  rqp[1] = ( get_chroma_qp( h, 1, mb_qp ) +
773  get_chroma_qp( h, 1, mbn1_qp ) + 1 ) >> 1;
774 
775  /* Filter edge */
776  tprintf(h->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);
777  { int i; for (i = 0; i < 8; i++) tprintf(h->avctx, " bS[%d]:%d", i, bS[i]); tprintf(h->avctx, "\n"); }
778  if(MB_FIELD){
779  filter_mb_mbaff_edgev ( h, img_y , linesize, bS , 1, qp [0], a, b, 1 );
780  filter_mb_mbaff_edgev ( h, img_y + 8* linesize, linesize, bS+4, 1, qp [1], a, b, 1 );
781  if (chroma){
782  if (CHROMA444) {
783  filter_mb_mbaff_edgev ( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 );
784  filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
785  filter_mb_mbaff_edgev ( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 );
786  filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
787  } else if (CHROMA422) {
788  filter_mb_mbaff_edgecv(h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1);
789  filter_mb_mbaff_edgecv(h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1);
790  filter_mb_mbaff_edgecv(h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1);
791  filter_mb_mbaff_edgecv(h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1);
792  }else{
793  filter_mb_mbaff_edgecv( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 );
794  filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
795  filter_mb_mbaff_edgecv( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 );
796  filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
797  }
798  }
799  }else{
800  filter_mb_mbaff_edgev ( h, img_y , 2* linesize, bS , 2, qp [0], a, b, 1 );
801  filter_mb_mbaff_edgev ( h, img_y + linesize, 2* linesize, bS+1, 2, qp [1], a, b, 1 );
802  if (chroma){
803  if (CHROMA444) {
804  filter_mb_mbaff_edgev ( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 );
805  filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
806  filter_mb_mbaff_edgev ( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 );
807  filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
808  }else{
809  filter_mb_mbaff_edgecv( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 );
810  filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
811  filter_mb_mbaff_edgecv( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 );
812  filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
813  }
814  }
815  }
816  }
817 
818 #if CONFIG_SMALL
819  for( dir = 0; dir < 2; dir++ )
820  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, a, b, chroma, dir);
821 #else
822  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, a, b, chroma, 0);
823  filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, a, b, chroma, 1);
824 #endif
825 }
#define LOCAL_ALIGNED_8(t, v,...)
Definition: dsputil.h:596
int cbp
Definition: h264.h:428
int left_type[LEFT_MBS]
Definition: h264.h:302
void(* h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:56
#define CHROMA422
Definition: h264.h:86
uint16_t * cbp_table
Definition: h264.h:427
int mb_y
Definition: h264.h:454
static const uint8_t alpha_table[52 *3]
void(* h264_h_loop_filter_luma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:60
int flags
Definition: h264.h:284
mpegvideo header.
void(* h264_h_loop_filter_chroma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:66
int16_t mv_cache[2][5 *8][2]
Motion vector cache.
Definition: h264.h:327
H264Context.
Definition: h264.h:252
int chroma_qp_diff
Definition: h264.h:224
int stride
Definition: mace.c:144
int picture_structure
Definition: h264.h:376
void(* h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:64
#define CONFIG_GRAY
Definition: config.h:277
#define IS_8x8DCT(a)
Definition: h264.h:94
uint8_t
#define PICT_FRAME
Definition: mpegvideo.h:629
void(* h264_h_loop_filter_luma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:53
#define b
Definition: input.c:52
int cabac
entropy_coding_mode_flag
Definition: h264.h:207
int mb_xy
Definition: h264.h:461
int uvlinesize
Definition: h264.h:274
int mb_x
Definition: h264.h:454
static av_always_inline void filter_mb_edgech(uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra)
int left_mb_xy[LEFT_MBS]
Definition: h264.h:297
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)
int top_mb_xy
Definition: h264.h:295
void(* h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:71
#define IS_INTERLACED(a)
Definition: mpegvideo.h:113
void(* h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:62
H.264 / AVC / MPEG4 part10 codec.
static av_always_inline void filter_mb_mbaff_edgecv(H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int a, int b, int intra)
static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale)
Get the chroma qp.
Definition: h264.h:815
PPS pps
current pps
Definition: h264.h:359
#define IS_INTRA(a)
Definition: mpegvideo.h:109
int off
Definition: dsputil_bfin.c:28
static const uint8_t scan8[16 *3+3]
Definition: h264.h:778
useful rectangle filling function
#define FRAME_MBAFF
Definition: h264.h:62
void(* h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:69
SPS sps
current sps
Definition: h264.h:354
#define MB_FIELD
Definition: h264.h:61
#define AV_WN64A(p, v)
Definition: intreadwrite.h:462
int top_type
Definition: h264.h:300
unsigned int list_count
Definition: h264.h:403
static av_always_inline void h264_filter_mb_fast_internal(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 pixel_shift)
Picture cur_pic
Definition: h264.h:264
uint32_t * mb_type
macroblock type table mb_type_base + mb_width + 2
Definition: avcodec.h:1180
int mb_stride
Definition: h264.h:459
AVCodecContext * avctx
Definition: h264.h:253
external API header
int slice_alpha_c0_offset
Definition: h264.h:467
#define LTOP
Definition: h264.h:65
void(* h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:49
static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit)
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 a, int b, int chroma, int dir)
static const uint8_t tc0_table[52 *3][4]
int slice_beta_offset
Definition: h264.h:468
static av_always_inline void filter_mb_mbaff_edgev(H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int a, int b, int intra)
int pixel_shift
0 for 8-bit H264, 1 for high-bit-depth H264
Definition: h264.h:268
static int step
Definition: avplay.c:252
#define CABAC
Definition: h264.h:83
static av_always_inline void filter_mb_edgeh(uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra)
void(* h264_h_loop_filter_chroma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:73
int8_t * qscale_table
QP table.
Definition: avcodec.h:1139
Definition: vf_drawbox.c:36
static av_always_inline void filter_mb_edgev(uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra)
#define AV_ZERO64(d)
Definition: intreadwrite.h:538
#define tprintf(p,...)
Definition: get_bits.h:613
common internal api header.
void(* h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:58
#define FIELD_PICTURE
Definition: h264.h:63
#define FILTER(hv, dir, edge, intra)
int chroma_qp[2]
Definition: h264.h:269
int bit_depth_luma
bit_depth_luma_minus8 + 8
Definition: h264.h:195
DSP utils.
int linesize
Definition: h264.h:274
void(* h264_loop_filter_strength)(int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2], int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field)
Definition: h264dsp.h:76
uint8_t non_zero_count_cache[15 *8]
non zero coeff count cache.
Definition: h264.h:320
void(* h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:51
struct AVFrame f
Definition: mpegvideo.h:96
H264DSPContext h264dsp
Definition: h264.h:256
static av_always_inline void filter_mb_edgecv(uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra)
#define LEFT
Definition: cdgraphics.c:171
uint8_t(* non_zero_count)[48]
Definition: h264.h:322
static const uint8_t beta_table[52 *3]
#define CHROMA444
Definition: h264.h:87
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)
int8_t ref_cache[2][5 *8]
Definition: h264.h:328