FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
h264idct_template.c
Go to the documentation of this file.
1 /*
2  * H.264 IDCT
3  * Copyright (c) 2004-2011 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 IDCT.
25  * @author Michael Niedermayer <michaelni@gmx.at>
26  */
27 
28 #include "bit_depth_template.c"
29 #include "libavutil/common.h"
30 
31 #ifndef AVCODEC_H264IDCT_INTERNAL_H
32 #define AVCODEC_H264IDCT_INTERNAL_H
33 //FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
34 static const uint8_t scan8[16*3]={
35  4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
36  6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
37  4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
38  6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
39  4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
40  6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
41  4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
42  6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
43  4+11*8, 5+11*8, 4+12*8, 5+12*8,
44  6+11*8, 7+11*8, 6+12*8, 7+12*8,
45  4+13*8, 5+13*8, 4+14*8, 5+14*8,
46  6+13*8, 7+13*8, 6+14*8, 7+14*8
47 };
48 #endif
49 
50 void FUNCC(ff_h264_idct_add)(uint8_t *_dst, DCTELEM *_block, int stride)
51 {
52  int i;
53  pixel *dst = (pixel*)_dst;
54  dctcoef *block = (dctcoef*)_block;
55  stride >>= sizeof(pixel)-1;
56 
57  block[0] += 1 << 5;
58 
59  for(i=0; i<4; i++){
60  const int z0= block[i + 4*0] + block[i + 4*2];
61  const int z1= block[i + 4*0] - block[i + 4*2];
62  const int z2= (block[i + 4*1]>>1) - block[i + 4*3];
63  const int z3= block[i + 4*1] + (block[i + 4*3]>>1);
64 
65  block[i + 4*0]= z0 + z3;
66  block[i + 4*1]= z1 + z2;
67  block[i + 4*2]= z1 - z2;
68  block[i + 4*3]= z0 - z3;
69  }
70 
71  for(i=0; i<4; i++){
72  const int z0= block[0 + 4*i] + block[2 + 4*i];
73  const int z1= block[0 + 4*i] - block[2 + 4*i];
74  const int z2= (block[1 + 4*i]>>1) - block[3 + 4*i];
75  const int z3= block[1 + 4*i] + (block[3 + 4*i]>>1);
76 
77  dst[i + 0*stride]= av_clip_pixel(dst[i + 0*stride] + ((z0 + z3) >> 6));
78  dst[i + 1*stride]= av_clip_pixel(dst[i + 1*stride] + ((z1 + z2) >> 6));
79  dst[i + 2*stride]= av_clip_pixel(dst[i + 2*stride] + ((z1 - z2) >> 6));
80  dst[i + 3*stride]= av_clip_pixel(dst[i + 3*stride] + ((z0 - z3) >> 6));
81  }
82 }
83 
84 void FUNCC(ff_h264_idct8_add)(uint8_t *_dst, DCTELEM *_block, int stride){
85  int i;
86  pixel *dst = (pixel*)_dst;
87  dctcoef *block = (dctcoef*)_block;
88  stride >>= sizeof(pixel)-1;
89 
90  block[0] += 32;
91 
92  for( i = 0; i < 8; i++ )
93  {
94  const int a0 = block[i+0*8] + block[i+4*8];
95  const int a2 = block[i+0*8] - block[i+4*8];
96  const int a4 = (block[i+2*8]>>1) - block[i+6*8];
97  const int a6 = (block[i+6*8]>>1) + block[i+2*8];
98 
99  const int b0 = a0 + a6;
100  const int b2 = a2 + a4;
101  const int b4 = a2 - a4;
102  const int b6 = a0 - a6;
103 
104  const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
105  const int a3 = block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
106  const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
107  const int a7 = block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
108 
109  const int b1 = (a7>>2) + a1;
110  const int b3 = a3 + (a5>>2);
111  const int b5 = (a3>>2) - a5;
112  const int b7 = a7 - (a1>>2);
113 
114  block[i+0*8] = b0 + b7;
115  block[i+7*8] = b0 - b7;
116  block[i+1*8] = b2 + b5;
117  block[i+6*8] = b2 - b5;
118  block[i+2*8] = b4 + b3;
119  block[i+5*8] = b4 - b3;
120  block[i+3*8] = b6 + b1;
121  block[i+4*8] = b6 - b1;
122  }
123  for( i = 0; i < 8; i++ )
124  {
125  const int a0 = block[0+i*8] + block[4+i*8];
126  const int a2 = block[0+i*8] - block[4+i*8];
127  const int a4 = (block[2+i*8]>>1) - block[6+i*8];
128  const int a6 = (block[6+i*8]>>1) + block[2+i*8];
129 
130  const int b0 = a0 + a6;
131  const int b2 = a2 + a4;
132  const int b4 = a2 - a4;
133  const int b6 = a0 - a6;
134 
135  const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
136  const int a3 = block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
137  const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
138  const int a7 = block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
139 
140  const int b1 = (a7>>2) + a1;
141  const int b3 = a3 + (a5>>2);
142  const int b5 = (a3>>2) - a5;
143  const int b7 = a7 - (a1>>2);
144 
145  dst[i + 0*stride] = av_clip_pixel( dst[i + 0*stride] + ((b0 + b7) >> 6) );
146  dst[i + 1*stride] = av_clip_pixel( dst[i + 1*stride] + ((b2 + b5) >> 6) );
147  dst[i + 2*stride] = av_clip_pixel( dst[i + 2*stride] + ((b4 + b3) >> 6) );
148  dst[i + 3*stride] = av_clip_pixel( dst[i + 3*stride] + ((b6 + b1) >> 6) );
149  dst[i + 4*stride] = av_clip_pixel( dst[i + 4*stride] + ((b6 - b1) >> 6) );
150  dst[i + 5*stride] = av_clip_pixel( dst[i + 5*stride] + ((b4 - b3) >> 6) );
151  dst[i + 6*stride] = av_clip_pixel( dst[i + 6*stride] + ((b2 - b5) >> 6) );
152  dst[i + 7*stride] = av_clip_pixel( dst[i + 7*stride] + ((b0 - b7) >> 6) );
153  }
154 }
155 
156 // assumes all AC coefs are 0
157 void FUNCC(ff_h264_idct_dc_add)(uint8_t *p_dst, DCTELEM *block, int stride){
158  int i, j;
159  int dc = (((dctcoef*)block)[0] + 32) >> 6;
160  pixel *dst = (pixel*)p_dst;
161  stride >>= sizeof(pixel)-1;
162  for( j = 0; j < 4; j++ )
163  {
164  for( i = 0; i < 4; i++ )
165  dst[i] = av_clip_pixel( dst[i] + dc );
166  dst += stride;
167  }
168 }
169 
170 void FUNCC(ff_h264_idct8_dc_add)(uint8_t *p_dst, DCTELEM *block, int stride){
171  int i, j;
172  int dc = (((dctcoef*)block)[0] + 32) >> 6;
173  pixel *dst = (pixel*)p_dst;
174  stride >>= sizeof(pixel)-1;
175  for( j = 0; j < 8; j++ )
176  {
177  for( i = 0; i < 8; i++ )
178  dst[i] = av_clip_pixel( dst[i] + dc );
179  dst += stride;
180  }
181 }
182 
183 void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
184  int i;
185  for(i=0; i<16; i++){
186  int nnz = nnzc[ scan8[i] ];
187  if(nnz){
188  if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
189  else FUNCC(ff_h264_idct_add )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
190  }
191  }
192 }
193 
194 void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
195  int i;
196  for(i=0; i<16; i++){
197  if(nnzc[ scan8[i] ]) FUNCC(ff_h264_idct_add )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
198  else if(((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
199  }
200 }
201 
202 void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
203  int i;
204  for(i=0; i<16; i+=4){
205  int nnz = nnzc[ scan8[i] ];
206  if(nnz){
207  if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct8_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
208  else FUNCC(ff_h264_idct8_add )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
209  }
210  }
211 }
212 
213 void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
214  int i, j;
215  for(j=1; j<3; j++){
216  for(i=j*16; i<j*16+4; i++){
217  if(nnzc[ scan8[i] ])
218  FUNCC(ff_h264_idct_add )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
219  else if(((dctcoef*)block)[i*16])
220  FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
221  }
222  }
223 }
224 
225 void FUNCC(ff_h264_idct_add8_422)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
226  int i, j;
227 
228  for(j=1; j<3; j++){
229  for(i=j*16; i<j*16+4; i++){
230  if(nnzc[ scan8[i] ])
231  FUNCC(ff_h264_idct_add )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
232  else if(((dctcoef*)block)[i*16])
233  FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
234  }
235  }
236 
237  for(j=1; j<3; j++){
238  for(i=j*16+4; i<j*16+8; i++){
239  if(nnzc[ scan8[i+4] ])
240  FUNCC(ff_h264_idct_add )(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
241  else if(((dctcoef*)block)[i*16])
242  FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
243  }
244  }
245 }
246 
247 /**
248  * IDCT transforms the 16 dc values and dequantizes them.
249  * @param qmul quantization parameter
250  */
251 void FUNCC(ff_h264_luma_dc_dequant_idct)(DCTELEM *p_output, DCTELEM *p_input, int qmul){
252 #define stride 16
253  int i;
254  int temp[16];
255  static const uint8_t x_offset[4]={0, 2*stride, 8*stride, 10*stride};
256  dctcoef *input = (dctcoef*)p_input;
257  dctcoef *output = (dctcoef*)p_output;
258 
259  for(i=0; i<4; i++){
260  const int z0= input[4*i+0] + input[4*i+1];
261  const int z1= input[4*i+0] - input[4*i+1];
262  const int z2= input[4*i+2] - input[4*i+3];
263  const int z3= input[4*i+2] + input[4*i+3];
264 
265  temp[4*i+0]= z0+z3;
266  temp[4*i+1]= z0-z3;
267  temp[4*i+2]= z1-z2;
268  temp[4*i+3]= z1+z2;
269  }
270 
271  for(i=0; i<4; i++){
272  const int offset= x_offset[i];
273  const int z0= temp[4*0+i] + temp[4*2+i];
274  const int z1= temp[4*0+i] - temp[4*2+i];
275  const int z2= temp[4*1+i] - temp[4*3+i];
276  const int z3= temp[4*1+i] + temp[4*3+i];
277 
278  output[stride* 0+offset]= ((((z0 + z3)*qmul + 128 ) >> 8));
279  output[stride* 1+offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
280  output[stride* 4+offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
281  output[stride* 5+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
282  }
283 #undef stride
284 }
285 
286 void FUNCC(ff_h264_chroma422_dc_dequant_idct)(DCTELEM *_block, int qmul){
287  const int stride= 16*2;
288  const int xStride= 16;
289  int i;
290  int temp[8];
291  static const uint8_t x_offset[2]={0, 16};
292  dctcoef *block = (dctcoef*)_block;
293 
294  for(i=0; i<4; i++){
295  temp[2*i+0] = block[stride*i + xStride*0] + block[stride*i + xStride*1];
296  temp[2*i+1] = block[stride*i + xStride*0] - block[stride*i + xStride*1];
297  }
298 
299  for(i=0; i<2; i++){
300  const int offset= x_offset[i];
301  const int z0= temp[2*0+i] + temp[2*2+i];
302  const int z1= temp[2*0+i] - temp[2*2+i];
303  const int z2= temp[2*1+i] - temp[2*3+i];
304  const int z3= temp[2*1+i] + temp[2*3+i];
305 
306  block[stride*0+offset]= ((z0 + z3)*qmul + 128) >> 8;
307  block[stride*1+offset]= ((z1 + z2)*qmul + 128) >> 8;
308  block[stride*2+offset]= ((z1 - z2)*qmul + 128) >> 8;
309  block[stride*3+offset]= ((z0 - z3)*qmul + 128) >> 8;
310  }
311 }
312 
313 void FUNCC(ff_h264_chroma_dc_dequant_idct)(DCTELEM *_block, int qmul){
314  const int stride= 16*2;
315  const int xStride= 16;
316  int a,b,c,d,e;
317  dctcoef *block = (dctcoef*)_block;
318 
319  a= block[stride*0 + xStride*0];
320  b= block[stride*0 + xStride*1];
321  c= block[stride*1 + xStride*0];
322  d= block[stride*1 + xStride*1];
323 
324  e= a-b;
325  a= a+b;
326  b= c-d;
327  c= c+d;
328 
329  block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
330  block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
331  block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
332  block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
333 }
Generated on Sat May 25 2013 03:58:35 for FFmpeg by   doxygen 1.8.2