blob: 0288cc761e10e43260f8535413934b15daa57cad
1 | /* |
2 | * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder |
3 | * Copyright (c) 2003-2011 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 DSP functions. |
25 | * @author Michael Niedermayer <michaelni@gmx.at> |
26 | */ |
27 | |
28 | #include "bit_depth_template.c" |
29 | |
30 | #define op_scale1(x) block[x] = av_clip_pixel( (block[x]*weight + offset) >> log2_denom ) |
31 | #define op_scale2(x) dst[x] = av_clip_pixel( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1)) |
32 | #define H264_WEIGHT(W) \ |
33 | static void FUNCC(weight_h264_pixels ## W)(uint8_t *_block, ptrdiff_t stride, int height, \ |
34 | int log2_denom, int weight, int offset) \ |
35 | { \ |
36 | int y; \ |
37 | pixel *block = (pixel*)_block; \ |
38 | stride >>= sizeof(pixel)-1; \ |
39 | offset = (unsigned)offset << (log2_denom + (BIT_DEPTH-8)); \ |
40 | if(log2_denom) offset += 1<<(log2_denom-1); \ |
41 | for (y = 0; y < height; y++, block += stride) { \ |
42 | op_scale1(0); \ |
43 | op_scale1(1); \ |
44 | if(W==2) continue; \ |
45 | op_scale1(2); \ |
46 | op_scale1(3); \ |
47 | if(W==4) continue; \ |
48 | op_scale1(4); \ |
49 | op_scale1(5); \ |
50 | op_scale1(6); \ |
51 | op_scale1(7); \ |
52 | if(W==8) continue; \ |
53 | op_scale1(8); \ |
54 | op_scale1(9); \ |
55 | op_scale1(10); \ |
56 | op_scale1(11); \ |
57 | op_scale1(12); \ |
58 | op_scale1(13); \ |
59 | op_scale1(14); \ |
60 | op_scale1(15); \ |
61 | } \ |
62 | } \ |
63 | static void FUNCC(biweight_h264_pixels ## W)(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride, int height, \ |
64 | int log2_denom, int weightd, int weights, int offset) \ |
65 | { \ |
66 | int y; \ |
67 | pixel *dst = (pixel*)_dst; \ |
68 | pixel *src = (pixel*)_src; \ |
69 | stride >>= sizeof(pixel)-1; \ |
70 | offset = (unsigned)offset << (BIT_DEPTH-8); \ |
71 | offset = (unsigned)((offset + 1) | 1) << log2_denom; \ |
72 | for (y = 0; y < height; y++, dst += stride, src += stride) { \ |
73 | op_scale2(0); \ |
74 | op_scale2(1); \ |
75 | if(W==2) continue; \ |
76 | op_scale2(2); \ |
77 | op_scale2(3); \ |
78 | if(W==4) continue; \ |
79 | op_scale2(4); \ |
80 | op_scale2(5); \ |
81 | op_scale2(6); \ |
82 | op_scale2(7); \ |
83 | if(W==8) continue; \ |
84 | op_scale2(8); \ |
85 | op_scale2(9); \ |
86 | op_scale2(10); \ |
87 | op_scale2(11); \ |
88 | op_scale2(12); \ |
89 | op_scale2(13); \ |
90 | op_scale2(14); \ |
91 | op_scale2(15); \ |
92 | } \ |
93 | } |
94 | |
95 | H264_WEIGHT(16) |
96 | H264_WEIGHT(8) |
97 | H264_WEIGHT(4) |
98 | H264_WEIGHT(2) |
99 | |
100 | #undef op_scale1 |
101 | #undef op_scale2 |
102 | #undef H264_WEIGHT |
103 | |
104 | static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0) |
105 | { |
106 | pixel *pix = (pixel*)p_pix; |
107 | int i, d; |
108 | xstride >>= sizeof(pixel)-1; |
109 | ystride >>= sizeof(pixel)-1; |
110 | alpha <<= BIT_DEPTH - 8; |
111 | beta <<= BIT_DEPTH - 8; |
112 | for( i = 0; i < 4; i++ ) { |
113 | const int tc_orig = tc0[i] * (1 << (BIT_DEPTH - 8)); |
114 | if( tc_orig < 0 ) { |
115 | pix += inner_iters*ystride; |
116 | continue; |
117 | } |
118 | for( d = 0; d < inner_iters; d++ ) { |
119 | const int p0 = pix[-1*xstride]; |
120 | const int p1 = pix[-2*xstride]; |
121 | const int p2 = pix[-3*xstride]; |
122 | const int q0 = pix[0]; |
123 | const int q1 = pix[1*xstride]; |
124 | const int q2 = pix[2*xstride]; |
125 | |
126 | if( FFABS( p0 - q0 ) < alpha && |
127 | FFABS( p1 - p0 ) < beta && |
128 | FFABS( q1 - q0 ) < beta ) { |
129 | |
130 | int tc = tc_orig; |
131 | int i_delta; |
132 | |
133 | if( FFABS( p2 - p0 ) < beta ) { |
134 | if(tc_orig) |
135 | pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc_orig, tc_orig ); |
136 | tc++; |
137 | } |
138 | if( FFABS( q2 - q0 ) < beta ) { |
139 | if(tc_orig) |
140 | pix[ xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc_orig, tc_orig ); |
141 | tc++; |
142 | } |
143 | |
144 | i_delta = av_clip( (((q0 - p0 ) * 4) + (p1 - q1) + 4) >> 3, -tc, tc ); |
145 | pix[-xstride] = av_clip_pixel( p0 + i_delta ); /* p0' */ |
146 | pix[0] = av_clip_pixel( q0 - i_delta ); /* q0' */ |
147 | } |
148 | pix += ystride; |
149 | } |
150 | } |
151 | } |
152 | static void FUNCC(h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
153 | { |
154 | FUNCC(h264_loop_filter_luma)(pix, stride, sizeof(pixel), 4, alpha, beta, tc0); |
155 | } |
156 | static void FUNCC(h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
157 | { |
158 | FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0); |
159 | } |
160 | static void FUNCC(h264_h_loop_filter_luma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
161 | { |
162 | FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0); |
163 | } |
164 | |
165 | static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma_intra)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta) |
166 | { |
167 | pixel *pix = (pixel*)p_pix; |
168 | int d; |
169 | xstride >>= sizeof(pixel)-1; |
170 | ystride >>= sizeof(pixel)-1; |
171 | alpha <<= BIT_DEPTH - 8; |
172 | beta <<= BIT_DEPTH - 8; |
173 | for( d = 0; d < 4 * inner_iters; d++ ) { |
174 | const int p2 = pix[-3*xstride]; |
175 | const int p1 = pix[-2*xstride]; |
176 | const int p0 = pix[-1*xstride]; |
177 | |
178 | const int q0 = pix[ 0*xstride]; |
179 | const int q1 = pix[ 1*xstride]; |
180 | const int q2 = pix[ 2*xstride]; |
181 | |
182 | if( FFABS( p0 - q0 ) < alpha && |
183 | FFABS( p1 - p0 ) < beta && |
184 | FFABS( q1 - q0 ) < beta ) { |
185 | |
186 | if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){ |
187 | if( FFABS( p2 - p0 ) < beta) |
188 | { |
189 | const int p3 = pix[-4*xstride]; |
190 | /* p0', p1', p2' */ |
191 | pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3; |
192 | pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2; |
193 | pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3; |
194 | } else { |
195 | /* p0' */ |
196 | pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; |
197 | } |
198 | if( FFABS( q2 - q0 ) < beta) |
199 | { |
200 | const int q3 = pix[3*xstride]; |
201 | /* q0', q1', q2' */ |
202 | pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3; |
203 | pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2; |
204 | pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3; |
205 | } else { |
206 | /* q0' */ |
207 | pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2; |
208 | } |
209 | }else{ |
210 | /* p0', q0' */ |
211 | pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; |
212 | pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2; |
213 | } |
214 | } |
215 | pix += ystride; |
216 | } |
217 | } |
218 | static void FUNCC(h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
219 | { |
220 | FUNCC(h264_loop_filter_luma_intra)(pix, stride, sizeof(pixel), 4, alpha, beta); |
221 | } |
222 | static void FUNCC(h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
223 | { |
224 | FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta); |
225 | } |
226 | static void FUNCC(h264_h_loop_filter_luma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta) |
227 | { |
228 | FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta); |
229 | } |
230 | |
231 | static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0) |
232 | { |
233 | pixel *pix = (pixel*)p_pix; |
234 | int i, d; |
235 | alpha <<= BIT_DEPTH - 8; |
236 | beta <<= BIT_DEPTH - 8; |
237 | xstride >>= sizeof(pixel)-1; |
238 | ystride >>= sizeof(pixel)-1; |
239 | for( i = 0; i < 4; i++ ) { |
240 | const int tc = ((tc0[i] - 1U) << (BIT_DEPTH - 8)) + 1; |
241 | if( tc <= 0 ) { |
242 | pix += inner_iters*ystride; |
243 | continue; |
244 | } |
245 | for( d = 0; d < inner_iters; d++ ) { |
246 | const int p0 = pix[-1*xstride]; |
247 | const int p1 = pix[-2*xstride]; |
248 | const int q0 = pix[0]; |
249 | const int q1 = pix[1*xstride]; |
250 | |
251 | if( FFABS( p0 - q0 ) < alpha && |
252 | FFABS( p1 - p0 ) < beta && |
253 | FFABS( q1 - q0 ) < beta ) { |
254 | |
255 | int delta = av_clip( ((q0 - p0) * 4 + (p1 - q1) + 4) >> 3, -tc, tc ); |
256 | |
257 | pix[-xstride] = av_clip_pixel( p0 + delta ); /* p0' */ |
258 | pix[0] = av_clip_pixel( q0 - delta ); /* q0' */ |
259 | } |
260 | pix += ystride; |
261 | } |
262 | } |
263 | } |
264 | static void FUNCC(h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
265 | { |
266 | FUNCC(h264_loop_filter_chroma)(pix, stride, sizeof(pixel), 2, alpha, beta, tc0); |
267 | } |
268 | static void FUNCC(h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
269 | { |
270 | FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0); |
271 | } |
272 | static void FUNCC(h264_h_loop_filter_chroma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
273 | { |
274 | FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 1, alpha, beta, tc0); |
275 | } |
276 | static void FUNCC(h264_h_loop_filter_chroma422)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
277 | { |
278 | FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0); |
279 | } |
280 | static void FUNCC(h264_h_loop_filter_chroma422_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
281 | { |
282 | FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0); |
283 | } |
284 | |
285 | static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma_intra)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta) |
286 | { |
287 | pixel *pix = (pixel*)p_pix; |
288 | int d; |
289 | xstride >>= sizeof(pixel)-1; |
290 | ystride >>= sizeof(pixel)-1; |
291 | alpha <<= BIT_DEPTH - 8; |
292 | beta <<= BIT_DEPTH - 8; |
293 | for( d = 0; d < 4 * inner_iters; d++ ) { |
294 | const int p0 = pix[-1*xstride]; |
295 | const int p1 = pix[-2*xstride]; |
296 | const int q0 = pix[0]; |
297 | const int q1 = pix[1*xstride]; |
298 | |
299 | if( FFABS( p0 - q0 ) < alpha && |
300 | FFABS( p1 - p0 ) < beta && |
301 | FFABS( q1 - q0 ) < beta ) { |
302 | |
303 | pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */ |
304 | pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */ |
305 | } |
306 | pix += ystride; |
307 | } |
308 | } |
309 | static void FUNCC(h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
310 | { |
311 | FUNCC(h264_loop_filter_chroma_intra)(pix, stride, sizeof(pixel), 2, alpha, beta); |
312 | } |
313 | static void FUNCC(h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
314 | { |
315 | FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta); |
316 | } |
317 | static void FUNCC(h264_h_loop_filter_chroma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta) |
318 | { |
319 | FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 1, alpha, beta); |
320 | } |
321 | static void FUNCC(h264_h_loop_filter_chroma422_intra)(uint8_t *pix, int stride, int alpha, int beta) |
322 | { |
323 | FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta); |
324 | } |
325 | static void FUNCC(h264_h_loop_filter_chroma422_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta) |
326 | { |
327 | FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta); |
328 | } |
329 |