blob: 0924f32613f1f06beaf7881c5c28d49f0cd62ba3
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 "h264dec.h" |
33 | #include "h264_ps.h" |
34 | #include "mathops.h" |
35 | #include "mpegutils.h" |
36 | #include "rectangle.h" |
37 | |
38 | /* Deblocking filter (p153) */ |
39 | static const uint8_t alpha_table[52*3] = { |
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, 0, 0, 0, |
44 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
45 | 0, 0, 0, 0, 0, 0, 4, 4, 5, 6, |
46 | 7, 8, 9, 10, 12, 13, 15, 17, 20, 22, |
47 | 25, 28, 32, 36, 40, 45, 50, 56, 63, 71, |
48 | 80, 90,101,113,127,144,162,182,203,226, |
49 | 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 | 255,255,255,255,255,255,255,255,255,255,255,255,255, |
54 | }; |
55 | static const uint8_t beta_table[52*3] = { |
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, 0, 0, 0, |
60 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
61 | 0, 0, 0, 0, 0, 0, 2, 2, 2, 3, |
62 | 3, 3, 3, 4, 4, 4, 6, 6, 7, 7, |
63 | 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, |
64 | 13, 13, 14, 14, 15, 15, 16, 16, 17, 17, |
65 | 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 | 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, |
70 | }; |
71 | static const uint8_t tc0_table[52*3][4] = { |
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 }, {-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 }, |
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, 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, 1 }, |
84 | {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 }, |
85 | {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, |
86 | {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 }, |
87 | {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 }, |
88 | {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 }, |
89 | {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-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 }, {-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 }, |
99 | }; |
100 | |
101 | /* intra: 0 if this loopfilter call is guaranteed to be inter (bS < 4), 1 if it might be intra (bS == 4) */ |
102 | static av_always_inline void filter_mb_edgev(uint8_t *pix, int stride, |
103 | const int16_t bS[4], |
104 | unsigned int qp, int a, int b, |
105 | const H264Context *h, int intra) |
106 | { |
107 | const unsigned int index_a = qp + a; |
108 | const int alpha = alpha_table[index_a]; |
109 | const int beta = beta_table[qp + b]; |
110 | if (alpha ==0 || beta == 0) return; |
111 | |
112 | if( bS[0] < 4 || !intra ) { |
113 | int8_t tc[4]; |
114 | tc[0] = tc0_table[index_a][bS[0]]; |
115 | tc[1] = tc0_table[index_a][bS[1]]; |
116 | tc[2] = tc0_table[index_a][bS[2]]; |
117 | tc[3] = tc0_table[index_a][bS[3]]; |
118 | h->h264dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc); |
119 | } else { |
120 | h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta); |
121 | } |
122 | } |
123 | |
124 | static av_always_inline void filter_mb_edgecv(uint8_t *pix, int stride, |
125 | const int16_t bS[4], |
126 | unsigned int qp, int a, int b, |
127 | const H264Context *h, int intra) |
128 | { |
129 | const unsigned int index_a = qp + a; |
130 | const int alpha = alpha_table[index_a]; |
131 | const int beta = beta_table[qp + b]; |
132 | if (alpha ==0 || beta == 0) return; |
133 | |
134 | if( bS[0] < 4 || !intra ) { |
135 | int8_t tc[4]; |
136 | tc[0] = tc0_table[index_a][bS[0]]+1; |
137 | tc[1] = tc0_table[index_a][bS[1]]+1; |
138 | tc[2] = tc0_table[index_a][bS[2]]+1; |
139 | tc[3] = tc0_table[index_a][bS[3]]+1; |
140 | h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc); |
141 | } else { |
142 | h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta); |
143 | } |
144 | } |
145 | |
146 | static av_always_inline void filter_mb_mbaff_edgev(const H264Context *h, uint8_t *pix, |
147 | int stride, |
148 | const int16_t bS[7], int bsi, |
149 | int qp, int a, int b, |
150 | int intra) |
151 | { |
152 | const unsigned int index_a = qp + a; |
153 | const int alpha = alpha_table[index_a]; |
154 | const int beta = beta_table[qp + b]; |
155 | if (alpha ==0 || beta == 0) return; |
156 | |
157 | if( bS[0] < 4 || !intra ) { |
158 | int8_t tc[4]; |
159 | tc[0] = tc0_table[index_a][bS[0*bsi]]; |
160 | tc[1] = tc0_table[index_a][bS[1*bsi]]; |
161 | tc[2] = tc0_table[index_a][bS[2*bsi]]; |
162 | tc[3] = tc0_table[index_a][bS[3*bsi]]; |
163 | h->h264dsp.h264_h_loop_filter_luma_mbaff(pix, stride, alpha, beta, tc); |
164 | } else { |
165 | h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta); |
166 | } |
167 | } |
168 | |
169 | static av_always_inline void filter_mb_mbaff_edgecv(const H264Context *h, |
170 | uint8_t *pix, int stride, |
171 | const int16_t bS[7], |
172 | int bsi, int qp, int a, |
173 | int b, int intra) |
174 | { |
175 | const unsigned int index_a = qp + a; |
176 | const int alpha = alpha_table[index_a]; |
177 | const int beta = beta_table[qp + b]; |
178 | if (alpha ==0 || beta == 0) return; |
179 | |
180 | if( bS[0] < 4 || !intra ) { |
181 | int8_t tc[4]; |
182 | tc[0] = tc0_table[index_a][bS[0*bsi]] + 1; |
183 | tc[1] = tc0_table[index_a][bS[1*bsi]] + 1; |
184 | tc[2] = tc0_table[index_a][bS[2*bsi]] + 1; |
185 | tc[3] = tc0_table[index_a][bS[3*bsi]] + 1; |
186 | h->h264dsp.h264_h_loop_filter_chroma_mbaff(pix, stride, alpha, beta, tc); |
187 | } else { |
188 | h->h264dsp.h264_h_loop_filter_chroma_mbaff_intra(pix, stride, alpha, beta); |
189 | } |
190 | } |
191 | |
192 | static av_always_inline void filter_mb_edgeh(uint8_t *pix, int stride, |
193 | const int16_t bS[4], |
194 | unsigned int qp, int a, int b, |
195 | const H264Context *h, int intra) |
196 | { |
197 | const unsigned int index_a = qp + a; |
198 | const int alpha = alpha_table[index_a]; |
199 | const int beta = beta_table[qp + b]; |
200 | if (alpha ==0 || beta == 0) return; |
201 | |
202 | if( bS[0] < 4 || !intra ) { |
203 | int8_t tc[4]; |
204 | tc[0] = tc0_table[index_a][bS[0]]; |
205 | tc[1] = tc0_table[index_a][bS[1]]; |
206 | tc[2] = tc0_table[index_a][bS[2]]; |
207 | tc[3] = tc0_table[index_a][bS[3]]; |
208 | h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc); |
209 | } else { |
210 | h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta); |
211 | } |
212 | } |
213 | |
214 | static av_always_inline void filter_mb_edgech(uint8_t *pix, int stride, |
215 | const int16_t bS[4], |
216 | unsigned int qp, int a, int b, |
217 | const H264Context *h, int intra) |
218 | { |
219 | const unsigned int index_a = qp + a; |
220 | const int alpha = alpha_table[index_a]; |
221 | const int beta = beta_table[qp + b]; |
222 | if (alpha ==0 || beta == 0) return; |
223 | |
224 | if( bS[0] < 4 || !intra ) { |
225 | int8_t tc[4]; |
226 | tc[0] = tc0_table[index_a][bS[0]]+1; |
227 | tc[1] = tc0_table[index_a][bS[1]]+1; |
228 | tc[2] = tc0_table[index_a][bS[2]]+1; |
229 | tc[3] = tc0_table[index_a][bS[3]]+1; |
230 | h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc); |
231 | } else { |
232 | h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta); |
233 | } |
234 | } |
235 | |
236 | static av_always_inline void h264_filter_mb_fast_internal(const H264Context *h, |
237 | H264SliceContext *sl, |
238 | int mb_x, int mb_y, |
239 | uint8_t *img_y, |
240 | uint8_t *img_cb, |
241 | uint8_t *img_cr, |
242 | unsigned int linesize, |
243 | unsigned int uvlinesize, |
244 | int pixel_shift) |
245 | { |
246 | int chroma = CHROMA(h) && !(CONFIG_GRAY && (h->flags & AV_CODEC_FLAG_GRAY)); |
247 | int chroma444 = CHROMA444(h); |
248 | int chroma422 = CHROMA422(h); |
249 | |
250 | int mb_xy = sl->mb_xy; |
251 | int left_type = sl->left_type[LTOP]; |
252 | int top_type = sl->top_type; |
253 | |
254 | int qp_bd_offset = 6 * (h->ps.sps->bit_depth_luma - 8); |
255 | int a = 52 + sl->slice_alpha_c0_offset - qp_bd_offset; |
256 | int b = 52 + sl->slice_beta_offset - qp_bd_offset; |
257 | |
258 | int mb_type = h->cur_pic.mb_type[mb_xy]; |
259 | int qp = h->cur_pic.qscale_table[mb_xy]; |
260 | int qp0 = h->cur_pic.qscale_table[mb_xy - 1]; |
261 | int qp1 = h->cur_pic.qscale_table[sl->top_mb_xy]; |
262 | int qpc = get_chroma_qp(h->ps.pps, 0, qp); |
263 | int qpc0 = get_chroma_qp(h->ps.pps, 0, qp0); |
264 | int qpc1 = get_chroma_qp(h->ps.pps, 0, qp1); |
265 | qp0 = (qp + qp0 + 1) >> 1; |
266 | qp1 = (qp + qp1 + 1) >> 1; |
267 | qpc0 = (qpc + qpc0 + 1) >> 1; |
268 | qpc1 = (qpc + qpc1 + 1) >> 1; |
269 | |
270 | if( IS_INTRA(mb_type) ) { |
271 | static const int16_t bS4[4] = {4,4,4,4}; |
272 | static const int16_t bS3[4] = {3,3,3,3}; |
273 | const int16_t *bSH = FIELD_PICTURE(h) ? bS3 : bS4; |
274 | if(left_type) |
275 | filter_mb_edgev( &img_y[4*0<<pixel_shift], linesize, bS4, qp0, a, b, h, 1); |
276 | if( IS_8x8DCT(mb_type) ) { |
277 | filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0); |
278 | if(top_type){ |
279 | filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1); |
280 | } |
281 | filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0); |
282 | } else { |
283 | filter_mb_edgev( &img_y[4*1<<pixel_shift], linesize, bS3, qp, a, b, h, 0); |
284 | filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0); |
285 | filter_mb_edgev( &img_y[4*3<<pixel_shift], linesize, bS3, qp, a, b, h, 0); |
286 | if(top_type){ |
287 | filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1); |
288 | } |
289 | filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, a, b, h, 0); |
290 | filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0); |
291 | filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, a, b, h, 0); |
292 | } |
293 | if(chroma){ |
294 | if(chroma444){ |
295 | if(left_type){ |
296 | filter_mb_edgev( &img_cb[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1); |
297 | filter_mb_edgev( &img_cr[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1); |
298 | } |
299 | if( IS_8x8DCT(mb_type) ) { |
300 | filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
301 | filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
302 | if(top_type){ |
303 | filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 ); |
304 | filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 ); |
305 | } |
306 | filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0); |
307 | filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0); |
308 | } else { |
309 | filter_mb_edgev( &img_cb[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
310 | filter_mb_edgev( &img_cr[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
311 | filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
312 | filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
313 | filter_mb_edgev( &img_cb[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
314 | filter_mb_edgev( &img_cr[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0); |
315 | if(top_type){ |
316 | filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1); |
317 | filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1); |
318 | } |
319 | filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, a, b, h, 0); |
320 | filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, a, b, h, 0); |
321 | filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0); |
322 | filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0); |
323 | filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, a, b, h, 0); |
324 | filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, a, b, h, 0); |
325 | } |
326 | }else if(chroma422){ |
327 | if(left_type){ |
328 | filter_mb_edgecv(&img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1); |
329 | filter_mb_edgecv(&img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1); |
330 | } |
331 | filter_mb_edgecv(&img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0); |
332 | filter_mb_edgecv(&img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0); |
333 | if(top_type){ |
334 | filter_mb_edgech(&img_cb[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1); |
335 | filter_mb_edgech(&img_cr[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1); |
336 | } |
337 | filter_mb_edgech(&img_cb[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
338 | filter_mb_edgech(&img_cr[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
339 | filter_mb_edgech(&img_cb[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
340 | filter_mb_edgech(&img_cr[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
341 | filter_mb_edgech(&img_cb[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
342 | filter_mb_edgech(&img_cr[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
343 | }else{ |
344 | if(left_type){ |
345 | filter_mb_edgecv( &img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1); |
346 | filter_mb_edgecv( &img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1); |
347 | } |
348 | filter_mb_edgecv( &img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0); |
349 | filter_mb_edgecv( &img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0); |
350 | if(top_type){ |
351 | filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1); |
352 | filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1); |
353 | } |
354 | filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
355 | filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0); |
356 | } |
357 | } |
358 | return; |
359 | } else { |
360 | LOCAL_ALIGNED(8, int16_t, bS, [2], [4][4]); |
361 | int edges; |
362 | if( IS_8x8DCT(mb_type) && (sl->cbp&7) == 7 && !chroma444 ) { |
363 | edges = 4; |
364 | AV_WN64A(bS[0][0], 0x0002000200020002ULL); |
365 | AV_WN64A(bS[0][2], 0x0002000200020002ULL); |
366 | AV_WN64A(bS[1][0], 0x0002000200020002ULL); |
367 | AV_WN64A(bS[1][2], 0x0002000200020002ULL); |
368 | } else { |
369 | 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; |
370 | 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; |
371 | int step = 1+(mb_type>>24); //IS_8x8DCT(mb_type) ? 2 : 1; |
372 | edges = 4 - 3*((mb_type>>3) & !(sl->cbp & 15)); //(mb_type & MB_TYPE_16x16) && !(h->cbp & 15) ? 1 : 4; |
373 | h->h264dsp.h264_loop_filter_strength(bS, sl->non_zero_count_cache, sl->ref_cache, sl->mv_cache, |
374 | sl->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE(h)); |
375 | } |
376 | if( IS_INTRA(left_type) ) |
377 | AV_WN64A(bS[0][0], 0x0004000400040004ULL); |
378 | if( IS_INTRA(top_type) ) |
379 | AV_WN64A(bS[1][0], FIELD_PICTURE(h) ? 0x0003000300030003ULL : 0x0004000400040004ULL); |
380 | |
381 | #define FILTER(hv,dir,edge,intra)\ |
382 | if(AV_RN64A(bS[dir][edge])) { \ |
383 | 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 );\ |
384 | if(chroma){\ |
385 | if(chroma444){\ |
386 | 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 );\ |
387 | 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 );\ |
388 | } else if(!(edge&1)) {\ |
389 | 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 );\ |
390 | 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 );\ |
391 | }\ |
392 | }\ |
393 | } |
394 | if(left_type) |
395 | FILTER(v,0,0,1); |
396 | if( edges == 1 ) { |
397 | if(top_type) |
398 | FILTER(h,1,0,1); |
399 | } else if( IS_8x8DCT(mb_type) ) { |
400 | FILTER(v,0,2,0); |
401 | if(top_type) |
402 | FILTER(h,1,0,1); |
403 | FILTER(h,1,2,0); |
404 | } else { |
405 | FILTER(v,0,1,0); |
406 | FILTER(v,0,2,0); |
407 | FILTER(v,0,3,0); |
408 | if(top_type) |
409 | FILTER(h,1,0,1); |
410 | FILTER(h,1,1,0); |
411 | FILTER(h,1,2,0); |
412 | FILTER(h,1,3,0); |
413 | } |
414 | #undef FILTER |
415 | } |
416 | } |
417 | |
418 | void ff_h264_filter_mb_fast(const H264Context *h, H264SliceContext *sl, |
419 | int mb_x, int mb_y, uint8_t *img_y, |
420 | uint8_t *img_cb, uint8_t *img_cr, |
421 | unsigned int linesize, unsigned int uvlinesize) |
422 | { |
423 | av_assert2(!FRAME_MBAFF(h)); |
424 | if(!h->h264dsp.h264_loop_filter_strength || h->ps.pps->chroma_qp_diff) { |
425 | ff_h264_filter_mb(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize); |
426 | return; |
427 | } |
428 | |
429 | #if CONFIG_SMALL |
430 | h264_filter_mb_fast_internal(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, h->pixel_shift); |
431 | #else |
432 | if(h->pixel_shift){ |
433 | h264_filter_mb_fast_internal(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 1); |
434 | }else{ |
435 | h264_filter_mb_fast_internal(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 0); |
436 | } |
437 | #endif |
438 | } |
439 | |
440 | static int check_mv(H264SliceContext *sl, long b_idx, long bn_idx, int mvy_limit) |
441 | { |
442 | int v; |
443 | |
444 | v = sl->ref_cache[0][b_idx] != sl->ref_cache[0][bn_idx]; |
445 | if (!v && sl->ref_cache[0][b_idx] != -1) |
446 | v = sl->mv_cache[0][b_idx][0] - sl->mv_cache[0][bn_idx][0] + 3 >= 7U | |
447 | FFABS(sl->mv_cache[0][b_idx][1] - sl->mv_cache[0][bn_idx][1]) >= mvy_limit; |
448 | |
449 | if (sl->list_count == 2) { |
450 | if(!v) |
451 | v = sl->ref_cache[1][b_idx] != sl->ref_cache[1][bn_idx] | |
452 | sl->mv_cache[1][b_idx][0] - sl->mv_cache[1][bn_idx][0] + 3 >= 7U | |
453 | FFABS(sl->mv_cache[1][b_idx][1] - sl->mv_cache[1][bn_idx][1]) >= mvy_limit; |
454 | |
455 | if(v){ |
456 | if (sl->ref_cache[0][b_idx] != sl->ref_cache[1][bn_idx] | |
457 | sl->ref_cache[1][b_idx] != sl->ref_cache[0][bn_idx]) |
458 | return 1; |
459 | return |
460 | sl->mv_cache[0][b_idx][0] - sl->mv_cache[1][bn_idx][0] + 3 >= 7U | |
461 | FFABS(sl->mv_cache[0][b_idx][1] - sl->mv_cache[1][bn_idx][1]) >= mvy_limit | |
462 | sl->mv_cache[1][b_idx][0] - sl->mv_cache[0][bn_idx][0] + 3 >= 7U | |
463 | FFABS(sl->mv_cache[1][b_idx][1] - sl->mv_cache[0][bn_idx][1]) >= mvy_limit; |
464 | } |
465 | } |
466 | |
467 | return v; |
468 | } |
469 | |
470 | static av_always_inline void filter_mb_dir(const H264Context *h, H264SliceContext *sl, |
471 | int mb_x, int mb_y, |
472 | uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, |
473 | unsigned int linesize, unsigned int uvlinesize, |
474 | int mb_xy, int mb_type, int mvy_limit, |
475 | int first_vertical_edge_done, int a, int b, |
476 | int chroma, int dir) |
477 | { |
478 | int edge; |
479 | int chroma_qp_avg[2]; |
480 | int chroma444 = CHROMA444(h); |
481 | int chroma422 = CHROMA422(h); |
482 | const int mbm_xy = dir == 0 ? mb_xy -1 : sl->top_mb_xy; |
483 | const int mbm_type = dir == 0 ? sl->left_type[LTOP] : sl->top_type; |
484 | |
485 | // how often to recheck mv-based bS when iterating between edges |
486 | static const uint8_t mask_edge_tab[2][8]={{0,3,3,3,1,1,1,1}, |
487 | {0,3,1,1,3,3,3,3}}; |
488 | const int mask_edge = mask_edge_tab[dir][(mb_type>>3)&7]; |
489 | const int edges = mask_edge== 3 && !(sl->cbp&15) ? 1 : 4; |
490 | |
491 | // how often to recheck mv-based bS when iterating along each edge |
492 | const int mask_par0 = mb_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)); |
493 | |
494 | if(mbm_type && !first_vertical_edge_done){ |
495 | |
496 | if (FRAME_MBAFF(h) && (dir == 1) && ((mb_y&1) == 0) |
497 | && IS_INTERLACED(mbm_type&~mb_type) |
498 | ) { |
499 | // This is a special case in the norm where the filtering must |
500 | // be done twice (one each of the field) even if we are in a |
501 | // frame macroblock. |
502 | // |
503 | unsigned int tmp_linesize = 2 * linesize; |
504 | unsigned int tmp_uvlinesize = 2 * uvlinesize; |
505 | int mbn_xy = mb_xy - 2 * h->mb_stride; |
506 | int j; |
507 | |
508 | for(j=0; j<2; j++, mbn_xy += h->mb_stride){ |
509 | LOCAL_ALIGNED(8, int16_t, bS, [4]); |
510 | int qp; |
511 | if (IS_INTRA(mb_type | h->cur_pic.mb_type[mbn_xy])) { |
512 | AV_WN64A(bS, 0x0003000300030003ULL); |
513 | } else { |
514 | if (!CABAC(h) && IS_8x8DCT(h->cur_pic.mb_type[mbn_xy])) { |
515 | bS[0]= 1+((h->cbp_table[mbn_xy] & 0x4000) || sl->non_zero_count_cache[scan8[0]+0]); |
516 | bS[1]= 1+((h->cbp_table[mbn_xy] & 0x4000) || sl->non_zero_count_cache[scan8[0]+1]); |
517 | bS[2]= 1+((h->cbp_table[mbn_xy] & 0x8000) || sl->non_zero_count_cache[scan8[0]+2]); |
518 | bS[3]= 1+((h->cbp_table[mbn_xy] & 0x8000) || sl->non_zero_count_cache[scan8[0]+3]); |
519 | }else{ |
520 | const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4; |
521 | int i; |
522 | for( i = 0; i < 4; i++ ) { |
523 | bS[i] = 1 + !!(sl->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]); |
524 | } |
525 | } |
526 | } |
527 | // Do not use s->qscale as luma quantizer because it has not the same |
528 | // value in IPCM macroblocks. |
529 | qp = (h->cur_pic.qscale_table[mb_xy] + h->cur_pic.qscale_table[mbn_xy] + 1) >> 1; |
530 | ff_tlog(h->avctx, "filter mb:%d/%d dir:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, qp, tmp_linesize, tmp_uvlinesize); |
531 | { int i; for (i = 0; i < 4; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); } |
532 | filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 ); |
533 | chroma_qp_avg[0] = (sl->chroma_qp[0] + get_chroma_qp(h->ps.pps, 0, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1; |
534 | chroma_qp_avg[1] = (sl->chroma_qp[1] + get_chroma_qp(h->ps.pps, 1, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1; |
535 | if (chroma) { |
536 | if (chroma444) { |
537 | filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0); |
538 | filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0); |
539 | } else { |
540 | filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0); |
541 | filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0); |
542 | } |
543 | } |
544 | } |
545 | }else{ |
546 | LOCAL_ALIGNED(8, int16_t, bS, [4]); |
547 | int qp; |
548 | |
549 | if( IS_INTRA(mb_type|mbm_type)) { |
550 | AV_WN64A(bS, 0x0003000300030003ULL); |
551 | if ( (!IS_INTERLACED(mb_type|mbm_type)) |
552 | || ((FRAME_MBAFF(h) || (h->picture_structure != PICT_FRAME)) && (dir == 0)) |
553 | ) |
554 | AV_WN64A(bS, 0x0004000400040004ULL); |
555 | } else { |
556 | int i; |
557 | int mv_done; |
558 | |
559 | if( dir && FRAME_MBAFF(h) && IS_INTERLACED(mb_type ^ mbm_type)) { |
560 | AV_WN64A(bS, 0x0001000100010001ULL); |
561 | mv_done = 1; |
562 | } |
563 | else if( mask_par0 && ((mbm_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) { |
564 | int b_idx= 8 + 4; |
565 | int bn_idx= b_idx - (dir ? 8:1); |
566 | |
567 | bS[0] = bS[1] = bS[2] = bS[3] = check_mv(sl, 8 + 4, bn_idx, mvy_limit); |
568 | mv_done = 1; |
569 | } |
570 | else |
571 | mv_done = 0; |
572 | |
573 | for( i = 0; i < 4; i++ ) { |
574 | int x = dir == 0 ? 0 : i; |
575 | int y = dir == 0 ? i : 0; |
576 | int b_idx= 8 + 4 + x + 8*y; |
577 | int bn_idx= b_idx - (dir ? 8:1); |
578 | |
579 | if (sl->non_zero_count_cache[b_idx] | |
580 | sl->non_zero_count_cache[bn_idx]) { |
581 | bS[i] = 2; |
582 | } |
583 | else if(!mv_done) |
584 | { |
585 | bS[i] = check_mv(sl, b_idx, bn_idx, mvy_limit); |
586 | } |
587 | } |
588 | } |
589 | |
590 | /* Filter edge */ |
591 | // Do not use s->qscale as luma quantizer because it has not the same |
592 | // value in IPCM macroblocks. |
593 | if(bS[0]+bS[1]+bS[2]+bS[3]){ |
594 | qp = (h->cur_pic.qscale_table[mb_xy] + h->cur_pic.qscale_table[mbm_xy] + 1) >> 1; |
595 | //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]); |
596 | 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); |
597 | //{ int i; for (i = 0; i < 4; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); } |
598 | chroma_qp_avg[0] = (sl->chroma_qp[0] + get_chroma_qp(h->ps.pps, 0, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1; |
599 | chroma_qp_avg[1] = (sl->chroma_qp[1] + get_chroma_qp(h->ps.pps, 1, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1; |
600 | if( dir == 0 ) { |
601 | filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 ); |
602 | if (chroma) { |
603 | if (chroma444) { |
604 | filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1); |
605 | filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1); |
606 | } else { |
607 | filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1); |
608 | filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1); |
609 | } |
610 | } |
611 | } else { |
612 | filter_mb_edgeh( &img_y[0], linesize, bS, qp, a, b, h, 1 ); |
613 | if (chroma) { |
614 | if (chroma444) { |
615 | filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1); |
616 | filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1); |
617 | } else { |
618 | filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1); |
619 | filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1); |
620 | } |
621 | } |
622 | } |
623 | } |
624 | } |
625 | } |
626 | |
627 | /* Calculate bS */ |
628 | for( edge = 1; edge < edges; edge++ ) { |
629 | LOCAL_ALIGNED(8, int16_t, bS, [4]); |
630 | int qp; |
631 | const int deblock_edge = !IS_8x8DCT(mb_type & (edge<<24)); // (edge&1) && IS_8x8DCT(mb_type) |
632 | |
633 | if (!deblock_edge && (!chroma422 || dir == 0)) |
634 | continue; |
635 | |
636 | if( IS_INTRA(mb_type)) { |
637 | AV_WN64A(bS, 0x0003000300030003ULL); |
638 | } else { |
639 | int i; |
640 | int mv_done; |
641 | |
642 | if( edge & mask_edge ) { |
643 | AV_ZERO64(bS); |
644 | mv_done = 1; |
645 | } |
646 | else if( mask_par0 ) { |
647 | int b_idx= 8 + 4 + edge * (dir ? 8:1); |
648 | int bn_idx= b_idx - (dir ? 8:1); |
649 | |
650 | bS[0] = bS[1] = bS[2] = bS[3] = check_mv(sl, b_idx, bn_idx, mvy_limit); |
651 | mv_done = 1; |
652 | } |
653 | else |
654 | mv_done = 0; |
655 | |
656 | for( i = 0; i < 4; i++ ) { |
657 | int x = dir == 0 ? edge : i; |
658 | int y = dir == 0 ? i : edge; |
659 | int b_idx= 8 + 4 + x + 8*y; |
660 | int bn_idx= b_idx - (dir ? 8:1); |
661 | |
662 | if (sl->non_zero_count_cache[b_idx] | |
663 | sl->non_zero_count_cache[bn_idx]) { |
664 | bS[i] = 2; |
665 | } |
666 | else if(!mv_done) |
667 | { |
668 | bS[i] = check_mv(sl, b_idx, bn_idx, mvy_limit); |
669 | } |
670 | } |
671 | |
672 | if(bS[0]+bS[1]+bS[2]+bS[3] == 0) |
673 | continue; |
674 | } |
675 | |
676 | /* Filter edge */ |
677 | // Do not use s->qscale as luma quantizer because it has not the same |
678 | // value in IPCM macroblocks. |
679 | qp = h->cur_pic.qscale_table[mb_xy]; |
680 | //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]); |
681 | 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); |
682 | //{ int i; for (i = 0; i < 4; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); } |
683 | if( dir == 0 ) { |
684 | filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 ); |
685 | if (chroma) { |
686 | if (chroma444) { |
687 | filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); |
688 | filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); |
689 | } else if( (edge&1) == 0 ) { |
690 | filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); |
691 | filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); |
692 | } |
693 | } |
694 | } else { |
695 | if (chroma422) { |
696 | if (deblock_edge) |
697 | filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0); |
698 | if (chroma) { |
699 | filter_mb_edgech(&img_cb[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); |
700 | filter_mb_edgech(&img_cr[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); |
701 | } |
702 | } else { |
703 | filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0); |
704 | if (chroma) { |
705 | if (chroma444) { |
706 | filter_mb_edgeh (&img_cb[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); |
707 | filter_mb_edgeh (&img_cr[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); |
708 | } else if ((edge&1) == 0) { |
709 | filter_mb_edgech(&img_cb[2*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); |
710 | filter_mb_edgech(&img_cr[2*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); |
711 | } |
712 | } |
713 | } |
714 | } |
715 | } |
716 | } |
717 | |
718 | void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, |
719 | int mb_x, int mb_y, |
720 | uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, |
721 | unsigned int linesize, unsigned int uvlinesize) |
722 | { |
723 | const int mb_xy= mb_x + mb_y*h->mb_stride; |
724 | const int mb_type = h->cur_pic.mb_type[mb_xy]; |
725 | const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4; |
726 | int first_vertical_edge_done = 0; |
727 | int chroma = CHROMA(h) && !(CONFIG_GRAY && (h->flags & AV_CODEC_FLAG_GRAY)); |
728 | int qp_bd_offset = 6 * (h->ps.sps->bit_depth_luma - 8); |
729 | int a = 52 + sl->slice_alpha_c0_offset - qp_bd_offset; |
730 | int b = 52 + sl->slice_beta_offset - qp_bd_offset; |
731 | |
732 | if (FRAME_MBAFF(h) |
733 | // and current and left pair do not have the same interlaced type |
734 | && IS_INTERLACED(mb_type ^ sl->left_type[LTOP]) |
735 | // and left mb is in available to us |
736 | && sl->left_type[LTOP]) { |
737 | /* First vertical edge is different in MBAFF frames |
738 | * There are 8 different bS to compute and 2 different Qp |
739 | */ |
740 | LOCAL_ALIGNED(8, int16_t, bS, [8]); |
741 | int qp[2]; |
742 | int bqp[2]; |
743 | int rqp[2]; |
744 | int mb_qp, mbn0_qp, mbn1_qp; |
745 | int i; |
746 | first_vertical_edge_done = 1; |
747 | |
748 | if( IS_INTRA(mb_type) ) { |
749 | AV_WN64A(&bS[0], 0x0004000400040004ULL); |
750 | AV_WN64A(&bS[4], 0x0004000400040004ULL); |
751 | } else { |
752 | static const uint8_t offset[2][2][8]={ |
753 | { |
754 | {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1}, |
755 | {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3}, |
756 | },{ |
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 | {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3}, |
759 | } |
760 | }; |
761 | const uint8_t *off= offset[MB_FIELD(sl)][mb_y&1]; |
762 | for( i = 0; i < 8; i++ ) { |
763 | int j= MB_FIELD(sl) ? i>>2 : i&1; |
764 | int mbn_xy = sl->left_mb_xy[LEFT(j)]; |
765 | int mbn_type = sl->left_type[LEFT(j)]; |
766 | |
767 | if( IS_INTRA( mbn_type ) ) |
768 | bS[i] = 4; |
769 | else{ |
770 | bS[i] = 1 + !!(sl->non_zero_count_cache[12+8*(i>>1)] | |
771 | ((!h->ps.pps->cabac && IS_8x8DCT(mbn_type)) ? |
772 | (h->cbp_table[mbn_xy] & (((MB_FIELD(sl) ? (i&2) : (mb_y&1)) ? 8 : 2) << 12)) |
773 | : |
774 | h->non_zero_count[mbn_xy][ off[i] ])); |
775 | } |
776 | } |
777 | } |
778 | |
779 | mb_qp = h->cur_pic.qscale_table[mb_xy]; |
780 | mbn0_qp = h->cur_pic.qscale_table[sl->left_mb_xy[0]]; |
781 | mbn1_qp = h->cur_pic.qscale_table[sl->left_mb_xy[1]]; |
782 | qp[0] = ( mb_qp + mbn0_qp + 1 ) >> 1; |
783 | bqp[0] = (get_chroma_qp(h->ps.pps, 0, mb_qp) + |
784 | get_chroma_qp(h->ps.pps, 0, mbn0_qp) + 1) >> 1; |
785 | rqp[0] = (get_chroma_qp(h->ps.pps, 1, mb_qp) + |
786 | get_chroma_qp(h->ps.pps, 1, mbn0_qp) + 1) >> 1; |
787 | qp[1] = ( mb_qp + mbn1_qp + 1 ) >> 1; |
788 | bqp[1] = (get_chroma_qp(h->ps.pps, 0, mb_qp) + |
789 | get_chroma_qp(h->ps.pps, 0, mbn1_qp) + 1 ) >> 1; |
790 | rqp[1] = (get_chroma_qp(h->ps.pps, 1, mb_qp) + |
791 | get_chroma_qp(h->ps.pps, 1, mbn1_qp) + 1 ) >> 1; |
792 | |
793 | /* Filter edge */ |
794 | 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); |
795 | { int i; for (i = 0; i < 8; i++) ff_tlog(h->avctx, " bS[%d]:%d", i, bS[i]); ff_tlog(h->avctx, "\n"); } |
796 | if (MB_FIELD(sl)) { |
797 | filter_mb_mbaff_edgev ( h, img_y , linesize, bS , 1, qp [0], a, b, 1 ); |
798 | filter_mb_mbaff_edgev ( h, img_y + 8* linesize, linesize, bS+4, 1, qp [1], a, b, 1 ); |
799 | if (chroma){ |
800 | if (CHROMA444(h)) { |
801 | filter_mb_mbaff_edgev ( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 ); |
802 | filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 ); |
803 | filter_mb_mbaff_edgev ( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 ); |
804 | filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 ); |
805 | } else if (CHROMA422(h)) { |
806 | filter_mb_mbaff_edgecv(h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1); |
807 | filter_mb_mbaff_edgecv(h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1); |
808 | filter_mb_mbaff_edgecv(h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1); |
809 | filter_mb_mbaff_edgecv(h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1); |
810 | }else{ |
811 | filter_mb_mbaff_edgecv( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 ); |
812 | filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 ); |
813 | filter_mb_mbaff_edgecv( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 ); |
814 | filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 ); |
815 | } |
816 | } |
817 | }else{ |
818 | filter_mb_mbaff_edgev ( h, img_y , 2* linesize, bS , 2, qp [0], a, b, 1 ); |
819 | filter_mb_mbaff_edgev ( h, img_y + linesize, 2* linesize, bS+1, 2, qp [1], a, b, 1 ); |
820 | if (chroma){ |
821 | if (CHROMA444(h)) { |
822 | filter_mb_mbaff_edgev ( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 ); |
823 | filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 ); |
824 | filter_mb_mbaff_edgev ( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 ); |
825 | filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 ); |
826 | }else{ |
827 | filter_mb_mbaff_edgecv( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 ); |
828 | filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 ); |
829 | filter_mb_mbaff_edgecv( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 ); |
830 | filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 ); |
831 | } |
832 | } |
833 | } |
834 | } |
835 | |
836 | #if CONFIG_SMALL |
837 | { |
838 | int dir; |
839 | for (dir = 0; dir < 2; dir++) |
840 | filter_mb_dir(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, |
841 | uvlinesize, mb_xy, mb_type, mvy_limit, |
842 | dir ? 0 : first_vertical_edge_done, a, b, |
843 | chroma, dir); |
844 | } |
845 | #else |
846 | 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); |
847 | 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); |
848 | #endif |
849 | } |
850 |