FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
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 FFmpeg.
6  *
7  * FFmpeg 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  * FFmpeg 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 FFmpeg; 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 / MPEG-4 part10 loop filter.
25  * @author Michael Niedermayer <michaelni@gmx.at>
26  */
27 
28 #include "libavutil/internal.h"
29 #include "libavutil/intreadwrite.h"
30 #include "internal.h"
31 #include "avcodec.h"
32 #include "h264.h"
33 #include "mathops.h"
34 #include "mpegutils.h"
35 #include "rectangle.h"
36 
37 /* Deblocking filter (p153) */
38 static const uint8_t alpha_table[52*3] = {
39  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
40  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
41  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
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,
44  0, 0, 0, 0, 0, 0, 4, 4, 5, 6,
45  7, 8, 9, 10, 12, 13, 15, 17, 20, 22,
46  25, 28, 32, 36, 40, 45, 50, 56, 63, 71,
47  80, 90,101,113,127,144,162,182,203,226,
48  255,255,
49  255,255,255,255,255,255,255,255,255,255,255,255,255,
50  255,255,255,255,255,255,255,255,255,255,255,255,255,
51  255,255,255,255,255,255,255,255,255,255,255,255,255,
52  255,255,255,255,255,255,255,255,255,255,255,255,255,
53 };
54 static const uint8_t beta_table[52*3] = {
55  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
56  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
57  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
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,
60  0, 0, 0, 0, 0, 0, 2, 2, 2, 3,
61  3, 3, 3, 4, 4, 4, 6, 6, 7, 7,
62  8, 8, 9, 9, 10, 10, 11, 11, 12, 12,
63  13, 13, 14, 14, 15, 15, 16, 16, 17, 17,
64  18, 18,
65  18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
66  18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
67  18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
68  18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
69 };
70 static const uint8_t tc0_table[52*3][4] = {
71  {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
72  {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
73  {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
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 },
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 }, {-1, 0, 0, 0 }, {-1, 0, 0, 1 },
83  {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 },
84  {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 },
85  {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 },
86  {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 },
87  {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 },
88  {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-1,13,17,25 },
89  {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
90  {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
91  {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-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 },
98 };
99 
100 /* intra: 0 if this loopfilter call is guaranteed to be inter (bS < 4), 1 if it might be intra (bS == 4) */
101 static av_always_inline void filter_mb_edgev(uint8_t *pix, int stride,
102  const int16_t bS[4],
103  unsigned int qp, int a, int b,
104  const H264Context *h, int intra)
105 {
106  const unsigned int index_a = qp + a;
107  const int alpha = alpha_table[index_a];
108  const int beta = beta_table[qp + b];
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 
123 static av_always_inline void filter_mb_edgecv(uint8_t *pix, int stride,
124  const int16_t bS[4],
125  unsigned int qp, int a, int b,
126  const H264Context *h, int intra)
127 {
128  const unsigned int index_a = qp + a;
129  const int alpha = alpha_table[index_a];
130  const int beta = beta_table[qp + b];
131  if (alpha ==0 || beta == 0) return;
132 
133  if( bS[0] < 4 || !intra ) {
134  int8_t tc[4];
135  tc[0] = tc0_table[index_a][bS[0]]+1;
136  tc[1] = tc0_table[index_a][bS[1]]+1;
137  tc[2] = tc0_table[index_a][bS[2]]+1;
138  tc[3] = tc0_table[index_a][bS[3]]+1;
139  h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);
140  } else {
141  h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);
142  }
143 }
144 
145 static av_always_inline void filter_mb_mbaff_edgev(const H264Context *h, uint8_t *pix,
146  int stride,
147  const int16_t bS[7], int bsi,
148  int qp, int a, int b,
149  int intra)
150 {
151  const unsigned int index_a = qp + a;
152  const int alpha = alpha_table[index_a];
153  const int beta = beta_table[qp + b];
154  if (alpha ==0 || beta == 0) return;
155 
156  if( bS[0] < 4 || !intra ) {
157  int8_t tc[4];
158  tc[0] = tc0_table[index_a][bS[0*bsi]];
159  tc[1] = tc0_table[index_a][bS[1*bsi]];
160  tc[2] = tc0_table[index_a][bS[2*bsi]];
161  tc[3] = tc0_table[index_a][bS[3*bsi]];
162  h->h264dsp.h264_h_loop_filter_luma_mbaff(pix, stride, alpha, beta, tc);
163  } else {
164  h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta);
165  }
166 }
167 
168 static av_always_inline void filter_mb_mbaff_edgecv(const H264Context *h,
169  uint8_t *pix, int stride,
170  const int16_t bS[7],
171  int bsi, int qp, int a,
172  int b, int intra)
173 {
174  const unsigned int index_a = qp + a;
175  const int alpha = alpha_table[index_a];
176  const int beta = beta_table[qp + b];
177  if (alpha ==0 || beta == 0) return;
178 
179  if( bS[0] < 4 || !intra ) {
180  int8_t tc[4];
181  tc[0] = tc0_table[index_a][bS[0*bsi]] + 1;
182  tc[1] = tc0_table[index_a][bS[1*bsi]] + 1;
183  tc[2] = tc0_table[index_a][bS[2*bsi]] + 1;
184  tc[3] = tc0_table[index_a][bS[3*bsi]] + 1;
185  h->h264dsp.h264_h_loop_filter_chroma_mbaff(pix, stride, alpha, beta, tc);
186  } else {
187  h->h264dsp.h264_h_loop_filter_chroma_mbaff_intra(pix, stride, alpha, beta);
188  }
189 }
190 
191 static av_always_inline void filter_mb_edgeh(uint8_t *pix, int stride,
192  const int16_t bS[4],
193  unsigned int qp, int a, int b,
194  const H264Context *h, int intra)
195 {
196  const unsigned int index_a = qp + a;
197  const int alpha = alpha_table[index_a];
198  const int beta = beta_table[qp + b];
199  if (alpha ==0 || beta == 0) return;
200 
201  if( bS[0] < 4 || !intra ) {
202  int8_t tc[4];
203  tc[0] = tc0_table[index_a][bS[0]];
204  tc[1] = tc0_table[index_a][bS[1]];
205  tc[2] = tc0_table[index_a][bS[2]];
206  tc[3] = tc0_table[index_a][bS[3]];
207  h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
208  } else {
209  h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);
210  }
211 }
212 
213 static av_always_inline void filter_mb_edgech(uint8_t *pix, int stride,
214  const int16_t bS[4],
215  unsigned int qp, int a, int b,
216  const H264Context *h, int intra)
217 {
218  const unsigned int index_a = qp + a;
219  const int alpha = alpha_table[index_a];
220  const int beta = beta_table[qp + b];
221  if (alpha ==0 || beta == 0) return;
222 
223  if( bS[0] < 4 || !intra ) {
224  int8_t tc[4];
225  tc[0] = tc0_table[index_a][bS[0]]+1;
226  tc[1] = tc0_table[index_a][bS[1]]+1;
227  tc[2] = tc0_table[index_a][bS[2]]+1;
228  tc[3] = tc0_table[index_a][bS[3]]+1;
229  h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);
230  } else {
231  h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);
232  }
233 }
234 
235 static av_always_inline void h264_filter_mb_fast_internal(const H264Context *h,
236  H264SliceContext *sl,
237  int mb_x, int mb_y,
238  uint8_t *img_y,
239  uint8_t *img_cb,
240  uint8_t *img_cr,
241  unsigned int linesize,
242  unsigned int uvlinesize,
243  int pixel_shift)
244 {
245  int chroma = CHROMA(h) && !(CONFIG_GRAY && (h->flags & AV_CODEC_FLAG_GRAY));
246  int chroma444 = CHROMA444(h);
247  int chroma422 = CHROMA422(h);
248 
249  int mb_xy = sl->mb_xy;
250  int left_type = sl->left_type[LTOP];
251  int top_type = sl->top_type;
252 
253  int qp_bd_offset = 6 * (h->ps.sps->bit_depth_luma - 8);
254  int a = 52 + sl->slice_alpha_c0_offset - qp_bd_offset;
255  int b = 52 + sl->slice_beta_offset - qp_bd_offset;
256 
257  int mb_type = h->cur_pic.mb_type[mb_xy];
258  int qp = h->cur_pic.qscale_table[mb_xy];
259  int qp0 = h->cur_pic.qscale_table[mb_xy - 1];
260  int qp1 = h->cur_pic.qscale_table[sl->top_mb_xy];
261  int qpc = get_chroma_qp( h, 0, qp );
262  int qpc0 = get_chroma_qp( h, 0, qp0 );
263  int qpc1 = get_chroma_qp( h, 0, qp1 );
264  qp0 = (qp + qp0 + 1) >> 1;
265  qp1 = (qp + qp1 + 1) >> 1;
266  qpc0 = (qpc + qpc0 + 1) >> 1;
267  qpc1 = (qpc + qpc1 + 1) >> 1;
268 
269  if( IS_INTRA(mb_type) ) {
270  static const int16_t bS4[4] = {4,4,4,4};
271  static const int16_t bS3[4] = {3,3,3,3};
272  const int16_t *bSH = FIELD_PICTURE(h) ? bS3 : bS4;
273  if(left_type)
274  filter_mb_edgev( &img_y[4*0<<pixel_shift], linesize, bS4, qp0, a, b, h, 1);
275  if( IS_8x8DCT(mb_type) ) {
276  filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
277  if(top_type){
278  filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
279  }
280  filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
281  } else {
282  filter_mb_edgev( &img_y[4*1<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
283  filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
284  filter_mb_edgev( &img_y[4*3<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
285  if(top_type){
286  filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
287  }
288  filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, a, b, h, 0);
289  filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
290  filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, a, b, h, 0);
291  }
292  if(chroma){
293  if(chroma444){
294  if(left_type){
295  filter_mb_edgev( &img_cb[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1);
296  filter_mb_edgev( &img_cr[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1);
297  }
298  if( IS_8x8DCT(mb_type) ) {
299  filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
300  filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
301  if(top_type){
302  filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
303  filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
304  }
305  filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
306  filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
307  } else {
308  filter_mb_edgev( &img_cb[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
309  filter_mb_edgev( &img_cr[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
310  filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
311  filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
312  filter_mb_edgev( &img_cb[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
313  filter_mb_edgev( &img_cr[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
314  if(top_type){
315  filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
316  filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
317  }
318  filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
319  filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
320  filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
321  filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
322  filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
323  filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
324  }
325  }else if(chroma422){
326  if(left_type){
327  filter_mb_edgecv(&img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
328  filter_mb_edgecv(&img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
329  }
330  filter_mb_edgecv(&img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
331  filter_mb_edgecv(&img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
332  if(top_type){
333  filter_mb_edgech(&img_cb[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
334  filter_mb_edgech(&img_cr[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
335  }
336  filter_mb_edgech(&img_cb[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
337  filter_mb_edgech(&img_cr[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
338  filter_mb_edgech(&img_cb[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
339  filter_mb_edgech(&img_cr[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
340  filter_mb_edgech(&img_cb[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
341  filter_mb_edgech(&img_cr[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
342  }else{
343  if(left_type){
344  filter_mb_edgecv( &img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
345  filter_mb_edgecv( &img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
346  }
347  filter_mb_edgecv( &img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
348  filter_mb_edgecv( &img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
349  if(top_type){
350  filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
351  filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
352  }
353  filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
354  filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
355  }
356  }
357  return;
358  } else {
359  LOCAL_ALIGNED(8, int16_t, bS, [2], [4][4]);
360  int edges;
361  if( IS_8x8DCT(mb_type) && (sl->cbp&7) == 7 && !chroma444 ) {
362  edges = 4;
363  AV_WN64A(bS[0][0], 0x0002000200020002ULL);
364  AV_WN64A(bS[0][2], 0x0002000200020002ULL);
365  AV_WN64A(bS[1][0], 0x0002000200020002ULL);
366  AV_WN64A(bS[1][2], 0x0002000200020002ULL);
367  } else {
368  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;
369  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;
370  int step = 1+(mb_type>>24); //IS_8x8DCT(mb_type) ? 2 : 1;
371  edges = 4 - 3*((mb_type>>3) & !(sl->cbp & 15)); //(mb_type & MB_TYPE_16x16) && !(h->cbp & 15) ? 1 : 4;
372  h->h264dsp.h264_loop_filter_strength(bS, sl->non_zero_count_cache, sl->ref_cache, sl->mv_cache,
373  sl->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE(h));
374  }
375  if( IS_INTRA(left_type) )
376  AV_WN64A(bS[0][0], 0x0004000400040004ULL);
377  if( IS_INTRA(top_type) )
378  AV_WN64A(bS[1][0], FIELD_PICTURE(h) ? 0x0003000300030003ULL : 0x0004000400040004ULL);
379 
380 #define FILTER(hv,dir,edge,intra)\
381  if(AV_RN64A(bS[dir][edge])) { \
382  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 );\
383  if(chroma){\
384  if(chroma444){\
385  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 );\
386  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 );\
387  } else if(!(edge&1)) {\
388  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 );\
389  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 );\
390  }\
391  }\
392  }
393  if(left_type)
394  FILTER(v,0,0,1);
395  if( edges == 1 ) {
396  if(top_type)
397  FILTER(h,1,0,1);
398  } else if( IS_8x8DCT(mb_type) ) {
399  FILTER(v,0,2,0);
400  if(top_type)
401  FILTER(h,1,0,1);
402  FILTER(h,1,2,0);
403  } else {
404  FILTER(v,0,1,0);
405  FILTER(v,0,2,0);
406  FILTER(v,0,3,0);
407  if(top_type)
408  FILTER(h,1,0,1);
409  FILTER(h,1,1,0);
410  FILTER(h,1,2,0);
411  FILTER(h,1,3,0);
412  }
413 #undef FILTER
414  }
415 }
416 
417 void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl,
418  int mb_x, int mb_y, uint8_t *img_y,
419  uint8_t *img_cb, uint8_t *img_cr,
420  unsigned int linesize, unsigned int uvlinesize)
421 {
422  av_assert2(!FRAME_MBAFF(h));
423  if(!h->h264dsp.h264_loop_filter_strength || h->ps.pps->chroma_qp_diff) {
424  ff_h264_filter_mb(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);
425  return;
426  }
427 
428 #if CONFIG_SMALL
429  h264_filter_mb_fast_internal(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, h->pixel_shift);
430 #else
431  if(h->pixel_shift){
432  h264_filter_mb_fast_internal(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 1);
433  }else{
434  h264_filter_mb_fast_internal(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 0);
435  }
436 #endif
437 }
438 
439 static int check_mv(H264SliceContext *sl, long b_idx, long bn_idx, int mvy_limit)
440 {
441  int v;
442 
443  v = sl->ref_cache[0][b_idx] != sl->ref_cache[0][bn_idx];
444  if (!v && sl->ref_cache[0][b_idx] != -1)
445  v = sl->mv_cache[0][b_idx][0] - sl->mv_cache[0][bn_idx][0] + 3 >= 7U |
446  FFABS(sl->mv_cache[0][b_idx][1] - sl->mv_cache[0][bn_idx][1]) >= mvy_limit;
447 
448  if (sl->list_count == 2) {
449  if(!v)
450  v = sl->ref_cache[1][b_idx] != sl->ref_cache[1][bn_idx] |
451  sl->mv_cache[1][b_idx][0] - sl->mv_cache[1][bn_idx][0] + 3 >= 7U |
452  FFABS(sl->mv_cache[1][b_idx][1] - sl->mv_cache[1][bn_idx][1]) >= mvy_limit;
453 
454  if(v){
455  if (sl->ref_cache[0][b_idx] != sl->ref_cache[1][bn_idx] |
456  sl->ref_cache[1][b_idx] != sl->ref_cache[0][bn_idx])
457  return 1;
458  return
459  sl->mv_cache[0][b_idx][0] - sl->mv_cache[1][bn_idx][0] + 3 >= 7U |
460  FFABS(sl->mv_cache[0][b_idx][1] - sl->mv_cache[1][bn_idx][1]) >= mvy_limit |
461  sl->mv_cache[1][b_idx][0] - sl->mv_cache[0][bn_idx][0] + 3 >= 7U |
462  FFABS(sl->mv_cache[1][b_idx][1] - sl->mv_cache[0][bn_idx][1]) >= mvy_limit;
463  }
464  }
465 
466  return v;
467 }
468 
469 static av_always_inline void filter_mb_dir(const H264Context *h, H264SliceContext *sl,
470  int mb_x, int mb_y,
471  uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
472  unsigned int linesize, unsigned int uvlinesize,
473  int mb_xy, int mb_type, int mvy_limit,
474  int first_vertical_edge_done, int a, int b,
475  int chroma, int dir)
476 {
477  int edge;
478  int chroma_qp_avg[2];
479  int chroma444 = CHROMA444(h);
480  int chroma422 = CHROMA422(h);
481  const int mbm_xy = dir == 0 ? mb_xy -1 : sl->top_mb_xy;
482  const int mbm_type = dir == 0 ? sl->left_type[LTOP] : sl->top_type;
483 
484  // how often to recheck mv-based bS when iterating between edges
485  static const uint8_t mask_edge_tab[2][8]={{0,3,3,3,1,1,1,1},
486  {0,3,1,1,3,3,3,3}};
487  const int mask_edge = mask_edge_tab[dir][(mb_type>>3)&7];
488  const int edges = mask_edge== 3 && !(sl->cbp&15) ? 1 : 4;
489 
490  // how often to recheck mv-based bS when iterating along each edge
491  const int mask_par0 = mb_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir));
492 
493  if(mbm_type && !first_vertical_edge_done){
494 
495  if (FRAME_MBAFF(h) && (dir == 1) && ((mb_y&1) == 0)
496  && IS_INTERLACED(mbm_type&~mb_type)
497  ) {
498  // This is a special case in the norm where the filtering must
499  // be done twice (one each of the field) even if we are in a
500  // frame macroblock.
501  //
502  unsigned int tmp_linesize = 2 * linesize;
503  unsigned int tmp_uvlinesize = 2 * uvlinesize;
504  int mbn_xy = mb_xy - 2 * h->mb_stride;
505  int j;
506 
507  for(j=0; j<2; j++, mbn_xy += h->mb_stride){
508  LOCAL_ALIGNED(8, int16_t, bS, [4]);
509  int qp;
510  if (IS_INTRA(mb_type | h->cur_pic.mb_type[mbn_xy])) {
511  AV_WN64A(bS, 0x0003000300030003ULL);
512  } else {
513  if (!CABAC(h) && IS_8x8DCT(h->cur_pic.mb_type[mbn_xy])) {
514  bS[0]= 1+((h->cbp_table[mbn_xy] & 0x4000) || sl->non_zero_count_cache[scan8[0]+0]);
515  bS[1]= 1+((h->cbp_table[mbn_xy] & 0x4000) || sl->non_zero_count_cache[scan8[0]+1]);
516  bS[2]= 1+((h->cbp_table[mbn_xy] & 0x8000) || sl->non_zero_count_cache[scan8[0]+2]);
517  bS[3]= 1+((h->cbp_table[mbn_xy] & 0x8000) || sl->non_zero_count_cache[scan8[0]+3]);
518  }else{
519  const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4;
520  int i;
521  for( i = 0; i < 4; i++ ) {
522  bS[i] = 1 + !!(sl->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]);
523  }
524  }
525  }
526  // Do not use s->qscale as luma quantizer because it has not the same
527  // value in IPCM macroblocks.
528  qp = (h->cur_pic.qscale_table[mb_xy] + h->cur_pic.qscale_table[mbn_xy] + 1) >> 1;
529  ff_tlog(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);
530  { int i; for (i = 0; i < 4; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); }
531  filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 );
532  chroma_qp_avg[0] = (sl->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1;
533  chroma_qp_avg[1] = (sl->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1;
534  if (chroma) {
535  if (chroma444) {
536  filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
537  filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
538  } else {
539  filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
540  filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
541  }
542  }
543  }
544  }else{
545  LOCAL_ALIGNED(8, int16_t, bS, [4]);
546  int qp;
547 
548  if( IS_INTRA(mb_type|mbm_type)) {
549  AV_WN64A(bS, 0x0003000300030003ULL);
550  if ( (!IS_INTERLACED(mb_type|mbm_type))
551  || ((FRAME_MBAFF(h) || (h->picture_structure != PICT_FRAME)) && (dir == 0))
552  )
553  AV_WN64A(bS, 0x0004000400040004ULL);
554  } else {
555  int i;
556  int mv_done;
557 
558  if( dir && FRAME_MBAFF(h) && IS_INTERLACED(mb_type ^ mbm_type)) {
559  AV_WN64A(bS, 0x0001000100010001ULL);
560  mv_done = 1;
561  }
562  else if( mask_par0 && ((mbm_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) {
563  int b_idx= 8 + 4;
564  int bn_idx= b_idx - (dir ? 8:1);
565 
566  bS[0] = bS[1] = bS[2] = bS[3] = check_mv(sl, 8 + 4, bn_idx, mvy_limit);
567  mv_done = 1;
568  }
569  else
570  mv_done = 0;
571 
572  for( i = 0; i < 4; i++ ) {
573  int x = dir == 0 ? 0 : i;
574  int y = dir == 0 ? i : 0;
575  int b_idx= 8 + 4 + x + 8*y;
576  int bn_idx= b_idx - (dir ? 8:1);
577 
578  if (sl->non_zero_count_cache[b_idx] |
579  sl->non_zero_count_cache[bn_idx]) {
580  bS[i] = 2;
581  }
582  else if(!mv_done)
583  {
584  bS[i] = check_mv(sl, b_idx, bn_idx, mvy_limit);
585  }
586  }
587  }
588 
589  /* Filter edge */
590  // Do not use s->qscale as luma quantizer because it has not the same
591  // value in IPCM macroblocks.
592  if(bS[0]+bS[1]+bS[2]+bS[3]){
593  qp = (h->cur_pic.qscale_table[mb_xy] + h->cur_pic.qscale_table[mbm_xy] + 1) >> 1;
594  //ff_tlog(h->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], h->cur_pic.qscale_table[mbn_xy]);
595  ff_tlog(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);
596  //{ int i; for (i = 0; i < 4; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); }
597  chroma_qp_avg[0] = (sl->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1;
598  chroma_qp_avg[1] = (sl->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1;
599  if( dir == 0 ) {
600  filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 );
601  if (chroma) {
602  if (chroma444) {
603  filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
604  filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
605  } else {
606  filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
607  filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
608  }
609  }
610  } else {
611  filter_mb_edgeh( &img_y[0], linesize, bS, qp, a, b, h, 1 );
612  if (chroma) {
613  if (chroma444) {
614  filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
615  filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
616  } else {
617  filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
618  filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
619  }
620  }
621  }
622  }
623  }
624  }
625 
626  /* Calculate bS */
627  for( edge = 1; edge < edges; edge++ ) {
628  LOCAL_ALIGNED(8, int16_t, bS, [4]);
629  int qp;
630  const int deblock_edge = !IS_8x8DCT(mb_type & (edge<<24)); // (edge&1) && IS_8x8DCT(mb_type)
631 
632  if (!deblock_edge && (!chroma422 || dir == 0))
633  continue;
634 
635  if( IS_INTRA(mb_type)) {
636  AV_WN64A(bS, 0x0003000300030003ULL);
637  } else {
638  int i;
639  int mv_done;
640 
641  if( edge & mask_edge ) {
642  AV_ZERO64(bS);
643  mv_done = 1;
644  }
645  else if( mask_par0 ) {
646  int b_idx= 8 + 4 + edge * (dir ? 8:1);
647  int bn_idx= b_idx - (dir ? 8:1);
648 
649  bS[0] = bS[1] = bS[2] = bS[3] = check_mv(sl, b_idx, bn_idx, mvy_limit);
650  mv_done = 1;
651  }
652  else
653  mv_done = 0;
654 
655  for( i = 0; i < 4; i++ ) {
656  int x = dir == 0 ? edge : i;
657  int y = dir == 0 ? i : edge;
658  int b_idx= 8 + 4 + x + 8*y;
659  int bn_idx= b_idx - (dir ? 8:1);
660 
661  if (sl->non_zero_count_cache[b_idx] |
662  sl->non_zero_count_cache[bn_idx]) {
663  bS[i] = 2;
664  }
665  else if(!mv_done)
666  {
667  bS[i] = check_mv(sl, b_idx, bn_idx, mvy_limit);
668  }
669  }
670 
671  if(bS[0]+bS[1]+bS[2]+bS[3] == 0)
672  continue;
673  }
674 
675  /* Filter edge */
676  // Do not use s->qscale as luma quantizer because it has not the same
677  // value in IPCM macroblocks.
678  qp = h->cur_pic.qscale_table[mb_xy];
679  //ff_tlog(h->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], h->cur_pic.qscale_table[mbn_xy]);
680  ff_tlog(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);
681  //{ int i; for (i = 0; i < 4; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); }
682  if( dir == 0 ) {
683  filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 );
684  if (chroma) {
685  if (chroma444) {
686  filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0);
687  filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0);
688  } else if( (edge&1) == 0 ) {
689  filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0);
690  filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0);
691  }
692  }
693  } else {
694  if (chroma422) {
695  if (deblock_edge)
696  filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
697  if (chroma) {
698  filter_mb_edgech(&img_cb[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0);
699  filter_mb_edgech(&img_cr[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0);
700  }
701  } else {
702  filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
703  if (chroma) {
704  if (chroma444) {
705  filter_mb_edgeh (&img_cb[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0);
706  filter_mb_edgeh (&img_cr[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0);
707  } else if ((edge&1) == 0) {
708  filter_mb_edgech(&img_cb[2*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0);
709  filter_mb_edgech(&img_cr[2*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0);
710  }
711  }
712  }
713  }
714  }
715 }
716 
717 void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl,
718  int mb_x, int mb_y,
719  uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr,
720  unsigned int linesize, unsigned int uvlinesize)
721 {
722  const int mb_xy= mb_x + mb_y*h->mb_stride;
723  const int mb_type = h->cur_pic.mb_type[mb_xy];
724  const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4;
725  int first_vertical_edge_done = 0;
726  int chroma = CHROMA(h) && !(CONFIG_GRAY && (h->flags & AV_CODEC_FLAG_GRAY));
727  int qp_bd_offset = 6 * (h->ps.sps->bit_depth_luma - 8);
728  int a = 52 + sl->slice_alpha_c0_offset - qp_bd_offset;
729  int b = 52 + sl->slice_beta_offset - qp_bd_offset;
730 
731  if (FRAME_MBAFF(h)
732  // and current and left pair do not have the same interlaced type
733  && IS_INTERLACED(mb_type ^ sl->left_type[LTOP])
734  // and left mb is in available to us
735  && sl->left_type[LTOP]) {
736  /* First vertical edge is different in MBAFF frames
737  * There are 8 different bS to compute and 2 different Qp
738  */
739  LOCAL_ALIGNED(8, int16_t, bS, [8]);
740  int qp[2];
741  int bqp[2];
742  int rqp[2];
743  int mb_qp, mbn0_qp, mbn1_qp;
744  int i;
745  first_vertical_edge_done = 1;
746 
747  if( IS_INTRA(mb_type) ) {
748  AV_WN64A(&bS[0], 0x0004000400040004ULL);
749  AV_WN64A(&bS[4], 0x0004000400040004ULL);
750  } else {
751  static const uint8_t offset[2][2][8]={
752  {
753  {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},
754  {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},
755  },{
756  {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
757  {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
758  }
759  };
760  const uint8_t *off= offset[MB_FIELD(sl)][mb_y&1];
761  for( i = 0; i < 8; i++ ) {
762  int j= MB_FIELD(sl) ? i>>2 : i&1;
763  int mbn_xy = sl->left_mb_xy[LEFT(j)];
764  int mbn_type = sl->left_type[LEFT(j)];
765 
766  if( IS_INTRA( mbn_type ) )
767  bS[i] = 4;
768  else{
769  bS[i] = 1 + !!(sl->non_zero_count_cache[12+8*(i>>1)] |
770  ((!h->ps.pps->cabac && IS_8x8DCT(mbn_type)) ?
771  (h->cbp_table[mbn_xy] & (((MB_FIELD(sl) ? (i&2) : (mb_y&1)) ? 8 : 2) << 12))
772  :
773  h->non_zero_count[mbn_xy][ off[i] ]));
774  }
775  }
776  }
777 
778  mb_qp = h->cur_pic.qscale_table[mb_xy];
779  mbn0_qp = h->cur_pic.qscale_table[sl->left_mb_xy[0]];
780  mbn1_qp = h->cur_pic.qscale_table[sl->left_mb_xy[1]];
781  qp[0] = ( mb_qp + mbn0_qp + 1 ) >> 1;
782  bqp[0] = ( get_chroma_qp( h, 0, mb_qp ) +
783  get_chroma_qp( h, 0, mbn0_qp ) + 1 ) >> 1;
784  rqp[0] = ( get_chroma_qp( h, 1, mb_qp ) +
785  get_chroma_qp( h, 1, mbn0_qp ) + 1 ) >> 1;
786  qp[1] = ( mb_qp + mbn1_qp + 1 ) >> 1;
787  bqp[1] = ( get_chroma_qp( h, 0, mb_qp ) +
788  get_chroma_qp( h, 0, mbn1_qp ) + 1 ) >> 1;
789  rqp[1] = ( get_chroma_qp( h, 1, mb_qp ) +
790  get_chroma_qp( h, 1, mbn1_qp ) + 1 ) >> 1;
791 
792  /* Filter edge */
793  ff_tlog(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);
794  { int i; for (i = 0; i < 8; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); }
795  if (MB_FIELD(sl)) {
796  filter_mb_mbaff_edgev ( h, img_y , linesize, bS , 1, qp [0], a, b, 1 );
797  filter_mb_mbaff_edgev ( h, img_y + 8* linesize, linesize, bS+4, 1, qp [1], a, b, 1 );
798  if (chroma){
799  if (CHROMA444(h)) {
800  filter_mb_mbaff_edgev ( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 );
801  filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
802  filter_mb_mbaff_edgev ( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 );
803  filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
804  } else if (CHROMA422(h)) {
805  filter_mb_mbaff_edgecv(h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1);
806  filter_mb_mbaff_edgecv(h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1);
807  filter_mb_mbaff_edgecv(h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1);
808  filter_mb_mbaff_edgecv(h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1);
809  }else{
810  filter_mb_mbaff_edgecv( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 );
811  filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
812  filter_mb_mbaff_edgecv( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 );
813  filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
814  }
815  }
816  }else{
817  filter_mb_mbaff_edgev ( h, img_y , 2* linesize, bS , 2, qp [0], a, b, 1 );
818  filter_mb_mbaff_edgev ( h, img_y + linesize, 2* linesize, bS+1, 2, qp [1], a, b, 1 );
819  if (chroma){
820  if (CHROMA444(h)) {
821  filter_mb_mbaff_edgev ( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 );
822  filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
823  filter_mb_mbaff_edgev ( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 );
824  filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
825  }else{
826  filter_mb_mbaff_edgecv( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 );
827  filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
828  filter_mb_mbaff_edgecv( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 );
829  filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
830  }
831  }
832  }
833  }
834 
835 #if CONFIG_SMALL
836  {
837  int dir;
838  for (dir = 0; dir < 2; dir++)
839  filter_mb_dir(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize,
840  uvlinesize, mb_xy, mb_type, mvy_limit,
841  dir ? 0 : first_vertical_edge_done, a, b,
842  chroma, dir);
843  }
844 #else
845  filter_mb_dir(h, sl, 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);
846  filter_mb_dir(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, a, b, chroma, 1);
847 #endif
848 }
ff_tlog
#define ff_tlog(ctx,...)
Definition: internal.h:65
chroma
static av_always_inline void chroma(WaveformContext *s, AVFrame *in, AVFrame *out, int component, int intensity, int offset_y, int offset_x, int column, int mirror)
Definition: vf_waveform.c:1318
check_mv
static int check_mv(H264SliceContext *sl, long b_idx, long bn_idx, int mvy_limit)
Definition: h264_loopfilter.c:439
h
h
Definition: vp9dsp_template.c:2093
H264SliceContext::mv_cache
int16_t mv_cache[2][5 *8][2]
Motion vector cache.
Definition: h264.h:429
CHROMA444
#define CHROMA444(h)
Definition: h264.h:103
H264SliceContext::left_mb_xy
int left_mb_xy[LEFT_MBS]
Definition: h264.h:350
H264DSPContext::h264_v_loop_filter_luma_intra
void(* h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:54
H264Context::cbp_table
uint16_t * cbp_table
Definition: h264.h:534
b
const char * b
Definition: vf_curves.c:109
alpha_table
static const uint8_t alpha_table[52 *3]
Definition: h264_loopfilter.c:38
H264DSPContext::h264_h_loop_filter_luma_mbaff_intra
void(* h264_h_loop_filter_luma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:58
H264Context::flags
int flags
Definition: h264.h:493
tc
#define tc
Definition: regdef.h:69
H264DSPContext::h264_h_loop_filter_chroma_mbaff
void(* h264_h_loop_filter_chroma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:64
H264Context
H264Context.
Definition: h264.h:456
PPS::chroma_qp_diff
int chroma_qp_diff
Definition: h264.h:219
H264Context::picture_structure
int picture_structure
Definition: h264.h:528
H264DSPContext::h264_h_loop_filter_chroma
void(* h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:62
IS_8x8DCT
#define IS_8x8DCT(a)
Definition: h264.h:110
MB_FIELD
#define MB_FIELD(sl)
Definition: h264.h:76
H264ParamSets::pps
const PPS * pps
Definition: h264.h:236
uint8_t
uint8_t
Definition: audio_convert.c:194
av_assert2
#define av_assert2(cond)
assert() equivalent, that does lie in speed critical code.
Definition: avassert.h:63
H264SliceContext::slice_alpha_c0_offset
int slice_alpha_c0_offset
Definition: h264.h:333
H264DSPContext::h264_h_loop_filter_luma_mbaff
void(* h264_h_loop_filter_luma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:51
PPS::cabac
int cabac
entropy_coding_mode_flag
Definition: h264.h:202
filter_mb_edgech
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, const H264Context *h, int intra)
Definition: h264_loopfilter.c:213
filter_mb_dir
static av_always_inline void filter_mb_dir(const H264Context *h, H264SliceContext *sl, 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)
Definition: h264_loopfilter.c:469
ff_h264_filter_mb
void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize)
Definition: h264_loopfilter.c:717
AV_CODEC_FLAG_GRAY
#define AV_CODEC_FLAG_GRAY
Only decode/encode grayscale.
Definition: avcodec.h:858
H264DSPContext::h264_h_loop_filter_chroma_intra
void(* h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:69
H264SliceContext::chroma_qp
int chroma_qp[2]
Definition: h264.h:327
H264DSPContext::h264_v_loop_filter_chroma
void(* h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:60
h264.h
H.264 / AVC / MPEG-4 part10 codec.
U
#define U(x)
Definition: vp56_arith.h:37
alpha
static double alpha(void *priv, double x, double y)
Definition: vf_geq.c:99
CHROMA
Definition: vf_waveform.c:36
CABAC
#define CABAC(h)
Definition: h264.h:98
offset
static const uint8_t offset[127][2]
Definition: vf_spp.c:92
scan8
static const uint8_t scan8[16 *3+3]
Definition: h264.h:801
internal.h
common internal API header
rectangle.h
useful rectangle filling function
FIELD_PICTURE
#define FIELD_PICTURE(h)
Definition: h264.h:78
H264DSPContext::h264_v_loop_filter_chroma_intra
void(* h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:67
filter_mb_edgeh
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, const H264Context *h, int intra)
Definition: h264_loopfilter.c:191
H264Picture::mb_type
uint32_t * mb_type
Definition: h264.h:274
H264SliceContext::top_mb_xy
int top_mb_xy
Definition: h264.h:348
FFABS
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:72
filter_mb_edgecv
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, const H264Context *h, int intra)
Definition: h264_loopfilter.c:123
AV_WN64A
#define AV_WN64A(p, v)
Definition: intreadwrite.h:542
filter_mb_mbaff_edgev
static av_always_inline void filter_mb_mbaff_edgev(const H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int a, int b, int intra)
Definition: h264_loopfilter.c:145
H264SliceContext::top_type
int top_type
Definition: h264.h:353
H264Context::mb_stride
int mb_stride
Definition: h264.h:556
IS_INTERLACED
#define IS_INTERLACED(a)
Definition: mpegutils.h:85
H264Context::avctx
AVCodecContext * avctx
Definition: h264.h:458
MB_TYPE_8x16
#define MB_TYPE_8x16
Definition: avcodec.h:1245
avcodec.h
Libavcodec external API header.
LTOP
#define LTOP
Definition: h264.h:80
h264_filter_mb_fast_internal
static av_always_inline void h264_filter_mb_fast_internal(const H264Context *h, H264SliceContext *sl, 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)
Definition: h264_loopfilter.c:235
H264DSPContext::h264_v_loop_filter_luma
void(* h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:47
H264Picture::qscale_table
int8_t * qscale_table
Definition: h264.h:268
tc0_table
static const uint8_t tc0_table[52 *3][4]
Definition: h264_loopfilter.c:70
H264SliceContext::non_zero_count_cache
uint8_t non_zero_count_cache[15 *8]
non zero coeff count cache.
Definition: h264.h:424
CHROMA422
#define CHROMA422(h)
Definition: h264.h:102
H264Context::pixel_shift
int pixel_shift
0 for 8-bit H.264, 1 for high-bit-depth H.264
Definition: h264.h:475
MB_TYPE_16x16
#define MB_TYPE_16x16
Definition: avcodec.h:1243
H264DSPContext::h264_h_loop_filter_chroma_mbaff_intra
void(* h264_h_loop_filter_chroma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:71
H264SliceContext::list_count
unsigned int list_count
Definition: h264.h:405
filter_mb_edgev
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, const H264Context *h, int intra)
Definition: h264_loopfilter.c:101
H264SliceContext::left_type
int left_type[LEFT_MBS]
Definition: h264.h:355
stride
GLint GLenum GLboolean GLsizei stride
Definition: opengl_enc.c:105
AV_ZERO64
#define AV_ZERO64(d)
Definition: intreadwrite.h:618
internal.h
common internal api header.
H264DSPContext::h264_h_loop_filter_luma_intra
void(* h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
Definition: h264dsp.h:56
LOCAL_ALIGNED
#define LOCAL_ALIGNED(a, t, v,...)
Definition: internal.h:110
FILTER
#define FILTER(hv, dir, edge, intra)
H264Context::ps
H264ParamSets ps
Definition: h264.h:574
SPS::bit_depth_luma
int bit_depth_luma
bit_depth_luma_minus8 + 8
Definition: h264.h:189
get_chroma_qp
static av_always_inline int get_chroma_qp(const H264Context *h, int t, int qscale)
Get the chroma qp.
Definition: h264.h:838
IS_INTRA
#define IS_INTRA(x, y)
PICT_FRAME
#define PICT_FRAME
Definition: mpegutils.h:39
H264DSPContext::h264_loop_filter_strength
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:74
H264SliceContext::ref_cache
int8_t ref_cache[2][5 *8]
Definition: h264.h:430
FRAME_MBAFF
#define FRAME_MBAFF(h)
Definition: h264.h:77
H264Context::cur_pic
H264Picture cur_pic
Definition: h264.h:467
H264DSPContext::h264_h_loop_filter_luma
void(* h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
Definition: h264dsp.h:49
filter_mb_mbaff_edgecv
static av_always_inline void filter_mb_mbaff_edgecv(const H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int a, int b, int intra)
Definition: h264_loopfilter.c:168
H264ParamSets::sps
SPS * sps
Definition: h264.h:238
H264Context::h264dsp
H264DSPContext h264dsp
Definition: h264.h:460
av_always_inline
#define av_always_inline
Definition: attributes.h:39
H264SliceContext::slice_beta_offset
int slice_beta_offset
Definition: h264.h:334
LEFT
#define LEFT
Definition: cdgraphics.c:168
H264Context::non_zero_count
uint8_t(* non_zero_count)[48]
Definition: h264.h:504
beta_table
static const uint8_t beta_table[52 *3]
Definition: h264_loopfilter.c:54
ff_h264_filter_mb_fast
void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize)
Definition: h264_loopfilter.c:417
a
a
Definition: h264pred_template.c:468
Generated on Mon Jun 27 2016 02:34:36 for FFmpeg by   doxygen 1.8.6