ffmpeg / libavcodec / h264dsp_internal.h @ 8dbe5856
History | View | Annotate | Download (9.84 KB)
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 / MPEG4 part10 DSP functions.
|
25 |
* @author Michael Niedermayer <michaelni@gmx.at>
|
26 |
*/
|
27 |
|
28 |
#include "h264_high_depth.h" |
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,H) \
|
33 |
static void FUNCC(weight_h264_pixels ## W ## x ## H)(uint8_t *p_block, int stride, int log2_denom, int weight, int offset){ \ |
34 |
int y; \
|
35 |
pixel *block = (pixel*)p_block; \ |
36 |
stride >>= sizeof(pixel)-1; \ |
37 |
offset <<= (log2_denom + (BIT_DEPTH-8)); \
|
38 |
if(log2_denom) offset += 1<<(log2_denom-1); \ |
39 |
for(y=0; y<H; y++, block += stride){ \ |
40 |
op_scale1(0); \
|
41 |
op_scale1(1); \
|
42 |
if(W==2) continue; \ |
43 |
op_scale1(2); \
|
44 |
op_scale1(3); \
|
45 |
if(W==4) continue; \ |
46 |
op_scale1(4); \
|
47 |
op_scale1(5); \
|
48 |
op_scale1(6); \
|
49 |
op_scale1(7); \
|
50 |
if(W==8) continue; \ |
51 |
op_scale1(8); \
|
52 |
op_scale1(9); \
|
53 |
op_scale1(10); \
|
54 |
op_scale1(11); \
|
55 |
op_scale1(12); \
|
56 |
op_scale1(13); \
|
57 |
op_scale1(14); \
|
58 |
op_scale1(15); \
|
59 |
} \ |
60 |
} \ |
61 |
static void FUNCC(biweight_h264_pixels ## W ## x ## H)(uint8_t *p_dst, uint8_t *p_src, int stride, int log2_denom, int weightd, int weights, int offset){ \ |
62 |
int y; \
|
63 |
pixel *dst = (pixel*)p_dst; \ |
64 |
pixel *src = (pixel*)p_src; \ |
65 |
stride >>= sizeof(pixel)-1; \ |
66 |
offset = ((offset + 1) | 1) << log2_denom; \ |
67 |
for(y=0; y<H; y++, dst += stride, src += stride){ \ |
68 |
op_scale2(0); \
|
69 |
op_scale2(1); \
|
70 |
if(W==2) continue; \ |
71 |
op_scale2(2); \
|
72 |
op_scale2(3); \
|
73 |
if(W==4) continue; \ |
74 |
op_scale2(4); \
|
75 |
op_scale2(5); \
|
76 |
op_scale2(6); \
|
77 |
op_scale2(7); \
|
78 |
if(W==8) continue; \ |
79 |
op_scale2(8); \
|
80 |
op_scale2(9); \
|
81 |
op_scale2(10); \
|
82 |
op_scale2(11); \
|
83 |
op_scale2(12); \
|
84 |
op_scale2(13); \
|
85 |
op_scale2(14); \
|
86 |
op_scale2(15); \
|
87 |
} \ |
88 |
} |
89 |
|
90 |
H264_WEIGHT(16,16) |
91 |
H264_WEIGHT(16,8) |
92 |
H264_WEIGHT(8,16) |
93 |
H264_WEIGHT(8,8) |
94 |
H264_WEIGHT(8,4) |
95 |
H264_WEIGHT(4,8) |
96 |
H264_WEIGHT(4,4) |
97 |
H264_WEIGHT(4,2) |
98 |
H264_WEIGHT(2,4) |
99 |
H264_WEIGHT(2,2) |
100 |
|
101 |
#undef op_scale1
|
102 |
#undef op_scale2
|
103 |
#undef H264_WEIGHT
|
104 |
|
105 |
static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma)(uint8_t *p_pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0) |
106 |
{ |
107 |
pixel *pix = (pixel*)p_pix; |
108 |
int i, d;
|
109 |
xstride >>= sizeof(pixel)-1; |
110 |
ystride >>= sizeof(pixel)-1; |
111 |
for( i = 0; i < 4; i++ ) { |
112 |
if( tc0[i] < 0 ) { |
113 |
pix += 4*ystride;
|
114 |
continue;
|
115 |
} |
116 |
for( d = 0; d < 4; d++ ) { |
117 |
const int p0 = pix[-1*xstride]; |
118 |
const int p1 = pix[-2*xstride]; |
119 |
const int p2 = pix[-3*xstride]; |
120 |
const int q0 = pix[0]; |
121 |
const int q1 = pix[1*xstride]; |
122 |
const int q2 = pix[2*xstride]; |
123 |
|
124 |
if( FFABS( p0 - q0 ) < alpha &&
|
125 |
FFABS( p1 - p0 ) < beta && |
126 |
FFABS( q1 - q0 ) < beta ) { |
127 |
|
128 |
int tc = tc0[i];
|
129 |
int i_delta;
|
130 |
|
131 |
if( FFABS( p2 - p0 ) < beta ) {
|
132 |
if(tc0[i])
|
133 |
pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] ); |
134 |
tc++; |
135 |
} |
136 |
if( FFABS( q2 - q0 ) < beta ) {
|
137 |
if(tc0[i])
|
138 |
pix[ xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] ); |
139 |
tc++; |
140 |
} |
141 |
|
142 |
i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc ); |
143 |
pix[-xstride] = av_clip_pixel( p0 + i_delta ); /* p0' */
|
144 |
pix[0] = av_clip_pixel( q0 - i_delta ); /* q0' */ |
145 |
} |
146 |
pix += ystride; |
147 |
} |
148 |
} |
149 |
} |
150 |
static void FUNCC(h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
151 |
{ |
152 |
FUNCC(h264_loop_filter_luma)(pix, stride, sizeof(pixel), alpha, beta, tc0);
|
153 |
} |
154 |
static void FUNCC(h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
155 |
{ |
156 |
FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, alpha, beta, tc0);
|
157 |
} |
158 |
|
159 |
static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma_intra)(uint8_t *p_pix, int xstride, int ystride, int alpha, int beta) |
160 |
{ |
161 |
pixel *pix = (pixel*)p_pix; |
162 |
int d;
|
163 |
xstride >>= sizeof(pixel)-1; |
164 |
ystride >>= sizeof(pixel)-1; |
165 |
for( d = 0; d < 16; d++ ) { |
166 |
const int p2 = pix[-3*xstride]; |
167 |
const int p1 = pix[-2*xstride]; |
168 |
const int p0 = pix[-1*xstride]; |
169 |
|
170 |
const int q0 = pix[ 0*xstride]; |
171 |
const int q1 = pix[ 1*xstride]; |
172 |
const int q2 = pix[ 2*xstride]; |
173 |
|
174 |
if( FFABS( p0 - q0 ) < alpha &&
|
175 |
FFABS( p1 - p0 ) < beta && |
176 |
FFABS( q1 - q0 ) < beta ) { |
177 |
|
178 |
if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){ |
179 |
if( FFABS( p2 - p0 ) < beta)
|
180 |
{ |
181 |
const int p3 = pix[-4*xstride]; |
182 |
/* p0', p1', p2' */
|
183 |
pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3; |
184 |
pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2; |
185 |
pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3; |
186 |
} else {
|
187 |
/* p0' */
|
188 |
pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; |
189 |
} |
190 |
if( FFABS( q2 - q0 ) < beta)
|
191 |
{ |
192 |
const int q3 = pix[3*xstride]; |
193 |
/* q0', q1', q2' */
|
194 |
pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3; |
195 |
pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2; |
196 |
pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3; |
197 |
} else {
|
198 |
/* q0' */
|
199 |
pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2; |
200 |
} |
201 |
}else{
|
202 |
/* p0', q0' */
|
203 |
pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; |
204 |
pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2; |
205 |
} |
206 |
} |
207 |
pix += ystride; |
208 |
} |
209 |
} |
210 |
static void FUNCC(h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
211 |
{ |
212 |
FUNCC(h264_loop_filter_luma_intra)(pix, stride, sizeof(pixel), alpha, beta);
|
213 |
} |
214 |
static void FUNCC(h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
215 |
{ |
216 |
FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, alpha, beta);
|
217 |
} |
218 |
|
219 |
static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma)(uint8_t *p_pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0) |
220 |
{ |
221 |
pixel *pix = (pixel*)p_pix; |
222 |
int i, d;
|
223 |
xstride >>= sizeof(pixel)-1; |
224 |
ystride >>= sizeof(pixel)-1; |
225 |
for( i = 0; i < 4; i++ ) { |
226 |
const int tc = tc0[i]; |
227 |
if( tc <= 0 ) { |
228 |
pix += 2*ystride;
|
229 |
continue;
|
230 |
} |
231 |
for( d = 0; d < 2; d++ ) { |
232 |
const int p0 = pix[-1*xstride]; |
233 |
const int p1 = pix[-2*xstride]; |
234 |
const int q0 = pix[0]; |
235 |
const int q1 = pix[1*xstride]; |
236 |
|
237 |
if( FFABS( p0 - q0 ) < alpha &&
|
238 |
FFABS( p1 - p0 ) < beta && |
239 |
FFABS( q1 - q0 ) < beta ) { |
240 |
|
241 |
int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc ); |
242 |
|
243 |
pix[-xstride] = av_clip_pixel( p0 + delta ); /* p0' */
|
244 |
pix[0] = av_clip_pixel( q0 - delta ); /* q0' */ |
245 |
} |
246 |
pix += ystride; |
247 |
} |
248 |
} |
249 |
} |
250 |
static void FUNCC(h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
251 |
{ |
252 |
FUNCC(h264_loop_filter_chroma)(pix, stride, sizeof(pixel), alpha, beta, tc0);
|
253 |
} |
254 |
static void FUNCC(h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) |
255 |
{ |
256 |
FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, alpha, beta, tc0);
|
257 |
} |
258 |
|
259 |
static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma_intra)(uint8_t *p_pix, int xstride, int ystride, int alpha, int beta) |
260 |
{ |
261 |
pixel *pix = (pixel*)p_pix; |
262 |
int d;
|
263 |
xstride >>= sizeof(pixel)-1; |
264 |
ystride >>= sizeof(pixel)-1; |
265 |
for( d = 0; d < 8; d++ ) { |
266 |
const int p0 = pix[-1*xstride]; |
267 |
const int p1 = pix[-2*xstride]; |
268 |
const int q0 = pix[0]; |
269 |
const int q1 = pix[1*xstride]; |
270 |
|
271 |
if( FFABS( p0 - q0 ) < alpha &&
|
272 |
FFABS( p1 - p0 ) < beta && |
273 |
FFABS( q1 - q0 ) < beta ) { |
274 |
|
275 |
pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */ |
276 |
pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */ |
277 |
} |
278 |
pix += ystride; |
279 |
} |
280 |
} |
281 |
static void FUNCC(h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
282 |
{ |
283 |
FUNCC(h264_loop_filter_chroma_intra)(pix, stride, sizeof(pixel), alpha, beta);
|
284 |
} |
285 |
static void FUNCC(h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta) |
286 |
{ |
287 |
FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, alpha, beta);
|
288 |
} |