1 |
/************************************************************************** |
/************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* decoder main |
* - Decoder main module - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* This program is an implementation of a part of one or more MPEG-4 |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
12 |
* editors and their companies, will have no liability for use of this |
* editors and their companies, will have no liability for use of this |
13 |
* software or modifications or derivatives thereof. |
* software or modifications or derivatives thereof. |
14 |
* |
* |
15 |
* This program is xvid_free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
16 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
17 |
* the xvid_free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
18 |
* (at your option) any later version. |
* (at your option) any later version. |
19 |
* |
* |
20 |
* This program is distributed in the hope that it will be useful, |
* This program is distributed in the hope that it will be useful, |
23 |
* GNU General Public License for more details. |
* GNU General Public License for more details. |
24 |
* |
* |
25 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
26 |
* along with this program; if not, write to the xvid_free Software |
* along with this program; if not, write to the Free Software |
27 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
28 |
* |
* |
29 |
*************************************************************************/ |
*************************************************************************/ |
30 |
|
|
32 |
* |
* |
33 |
* History: |
* History: |
34 |
* |
* |
35 |
|
* 22.06.2002 added primative N_VOP support |
36 |
|
* #define BFRAMES_DEC now enables Minchenm's bframe decoder |
37 |
|
* 08.05.2002 add low_delay support for B_VOP decode |
38 |
|
* MinChen <chenm001@163.com> |
39 |
|
* 05.05.2002 fix some B-frame decode problem |
40 |
|
* 02.05.2002 add B-frame decode support(have some problem); |
41 |
|
* MinChen <chenm001@163.com> |
42 |
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
43 |
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
44 |
* reconstructing blocks, thus artifacts |
* reconstructing blocks, thus artifacts |
46 |
* interlaced decoding should be as fast as progressive now |
* interlaced decoding should be as fast as progressive now |
47 |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
48 |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
49 |
* 22.12.2001 block based interpolation |
* 22.12.2001 lock based interpolation |
50 |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
51 |
* |
* |
52 |
|
* $Id: decoder.c,v 1.25 2002-07-09 01:09:33 chenm001 Exp $ |
53 |
|
* |
54 |
*************************************************************************/ |
*************************************************************************/ |
55 |
|
|
56 |
#include <stdlib.h> |
#include <stdlib.h> |
57 |
#include <string.h> // memset |
#include <string.h> |
58 |
|
|
59 |
#include "xvid.h" |
#include "xvid.h" |
60 |
#include "portab.h" |
#include "portab.h" |
79 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
80 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
81 |
|
|
82 |
int decoder_create(XVID_DEC_PARAM * param) |
int |
83 |
|
decoder_create(XVID_DEC_PARAM * param) |
84 |
{ |
{ |
85 |
DECODER * dec; |
DECODER * dec; |
86 |
|
|
87 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
88 |
if (dec == NULL) |
if (dec == NULL) { |
|
{ |
|
89 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
90 |
} |
} |
91 |
param->handle = dec; |
param->handle = dec; |
98 |
|
|
99 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
100 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
101 |
|
dec->low_delay = 0; |
102 |
|
|
103 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
|
{ |
|
104 |
xvid_free(dec); |
xvid_free(dec); |
105 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
106 |
} |
} |
107 |
|
|
108 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
109 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
110 |
xvid_free(dec); |
xvid_free(dec); |
111 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
112 |
} |
} |
113 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
114 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
115 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
116 |
{ |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
117 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
118 |
|
xvid_free(dec); |
119 |
|
return XVID_ERR_MEMORY; |
120 |
|
} |
121 |
|
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
122 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
123 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
124 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
125 |
xvid_free(dec); |
xvid_free(dec); |
126 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
127 |
} |
} |
128 |
|
|
129 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
dec->mbs = |
130 |
if (dec->mbs == NULL) |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
131 |
{ |
CACHE_LINE); |
132 |
|
if (dec->mbs == NULL) { |
133 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
134 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
135 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
136 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
137 |
xvid_free(dec); |
xvid_free(dec); |
138 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
139 |
} |
} |
140 |
|
|
141 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
142 |
|
|
143 |
|
// add by chenm001 <chenm001@163.com> |
144 |
|
// for skip MB flag |
145 |
|
dec->last_mbs = |
146 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
147 |
|
CACHE_LINE); |
148 |
|
if (dec->last_mbs == NULL) { |
149 |
|
xvid_free(dec->mbs); |
150 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
151 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
152 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
153 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
154 |
|
xvid_free(dec); |
155 |
|
return XVID_ERR_MEMORY; |
156 |
|
} |
157 |
|
|
158 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
159 |
|
|
160 |
init_timer(); |
init_timer(); |
161 |
|
|
162 |
|
// add by chenm001 <chenm001@163.com> |
163 |
|
// for support B-frame to save reference frame's time |
164 |
|
dec->frames = -1; |
165 |
|
dec->time = dec->time_base = dec->last_time_base = 0; |
166 |
|
|
167 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
168 |
} |
} |
169 |
|
|
170 |
|
|
171 |
int decoder_destroy(DECODER * dec) |
|
172 |
|
/* **************************************************************** |
173 |
|
NIC 28.06.2002 |
174 |
|
riscritta la 'decoder_create' per cambiarne l'interfaccia e poterla |
175 |
|
usare nella dll caricata dall'IM1 player |
176 |
|
IM1_decoder_create(XVID_DEC_PARAM * param,XVID_DEC_FRAME * frame) |
177 |
|
**************************************************************** */ |
178 |
|
//XVID_DEC_FRAME * frame |
179 |
|
|
180 |
|
int |
181 |
|
IM1_decoder_create(XVID_DEC_PARAM * param,XVID_DEC_FRAME * frame) |
182 |
|
{ |
183 |
|
DECODER *dec; |
184 |
|
Bitstream bs; |
185 |
|
uint32_t rounding; |
186 |
|
uint32_t quant; |
187 |
|
uint32_t fcode_forward; |
188 |
|
uint32_t fcode_backward; |
189 |
|
uint32_t intra_dc_threshold; |
190 |
|
|
191 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
192 |
|
if (dec == NULL) { |
193 |
|
return XVID_ERR_MEMORY; |
194 |
|
} |
195 |
|
param->handle = dec; |
196 |
|
|
197 |
|
|
198 |
|
BitstreamInit(&bs, frame->bitstream, frame->length);//NIC |
199 |
|
|
200 |
|
//NIC |
201 |
|
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
202 |
|
&fcode_backward, &intra_dc_threshold); |
203 |
|
|
204 |
|
|
205 |
|
param->width = dec->width; //NIC added |
206 |
|
param->height = dec->height; //NIC added |
207 |
|
//dec->width = param->width; //NIC commentate |
208 |
|
//dec->height = param->height; //NIC commentate |
209 |
|
|
210 |
|
dec->mb_width = (dec->width + 15) / 16; |
211 |
|
dec->mb_height = (dec->height + 15) / 16; |
212 |
|
|
213 |
|
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
214 |
|
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
215 |
|
dec->low_delay = 0; |
216 |
|
|
217 |
|
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
218 |
|
xvid_free(dec); |
219 |
|
return XVID_ERR_MEMORY; |
220 |
|
} |
221 |
|
|
222 |
|
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
223 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
224 |
|
xvid_free(dec); |
225 |
|
return XVID_ERR_MEMORY; |
226 |
|
} |
227 |
|
// add by chenm001 <chenm001@163.com> |
228 |
|
// for support B-frame to reference last 2 frame |
229 |
|
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
230 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
231 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
232 |
|
xvid_free(dec); |
233 |
|
return XVID_ERR_MEMORY; |
234 |
|
} |
235 |
|
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
236 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
237 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
238 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
239 |
|
xvid_free(dec); |
240 |
|
return XVID_ERR_MEMORY; |
241 |
|
} |
242 |
|
|
243 |
|
dec->mbs = |
244 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
245 |
|
CACHE_LINE); |
246 |
|
if (dec->mbs == NULL) { |
247 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
248 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
249 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
250 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
251 |
|
xvid_free(dec); |
252 |
|
return XVID_ERR_MEMORY; |
253 |
|
} |
254 |
|
|
255 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
256 |
|
|
257 |
|
// add by chenm001 <chenm001@163.com> |
258 |
|
// for skip MB flag |
259 |
|
dec->last_mbs = |
260 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
261 |
|
CACHE_LINE); |
262 |
|
if (dec->last_mbs == NULL) { |
263 |
|
xvid_free(dec->mbs); |
264 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
265 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
266 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
267 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
268 |
|
xvid_free(dec); |
269 |
|
return XVID_ERR_MEMORY; |
270 |
|
} |
271 |
|
|
272 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
273 |
|
|
274 |
|
init_timer(); |
275 |
|
|
276 |
|
// add by chenm001 <chenm001@163.com> |
277 |
|
// for support B-frame to save reference frame's time |
278 |
|
dec->frames = -1; |
279 |
|
dec->time = dec->time_base = dec->last_time_base = 0; |
280 |
|
|
281 |
|
return XVID_ERR_OK; |
282 |
|
} |
283 |
|
/* **************************************************************** |
284 |
|
END NIC |
285 |
|
**************************************************************** */ |
286 |
|
|
287 |
|
int |
288 |
|
decoder_destroy(DECODER * dec) |
289 |
{ |
{ |
290 |
|
xvid_free(dec->last_mbs); |
291 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
292 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
293 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
294 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
295 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
296 |
xvid_free(dec); |
xvid_free(dec); |
297 |
|
|
301 |
|
|
302 |
|
|
303 |
|
|
304 |
static const int32_t dquant_table[4] = |
static const int32_t dquant_table[4] = { |
|
{ |
|
305 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
306 |
}; |
}; |
307 |
|
|
308 |
|
|
309 |
|
|
310 |
|
|
311 |
// decode an intra macroblock |
// decode an intra macroblock |
312 |
|
|
313 |
void decoder_mbintra(DECODER * dec, |
void |
314 |
|
decoder_mbintra(DECODER * dec, |
315 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
316 |
const uint32_t x_pos, |
const uint32_t x_pos, |
317 |
const uint32_t y_pos, |
const uint32_t y_pos, |
338 |
|
|
339 |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
340 |
|
|
341 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
342 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
343 |
int16_t predictors[8]; |
int16_t predictors[8]; |
344 |
int start_coeff; |
int start_coeff; |
345 |
|
|
346 |
start_timer(); |
start_timer(); |
347 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
348 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors); |
349 |
{ |
if (!acpred_flag) { |
350 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
351 |
} |
} |
352 |
stop_prediction_timer(); |
stop_prediction_timer(); |
353 |
|
|
354 |
if (quant < intra_dc_threshold) |
if (quant < intra_dc_threshold) { |
|
{ |
|
355 |
int dc_size; |
int dc_size; |
356 |
int dc_dif; |
int dc_dif; |
357 |
|
|
358 |
dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); |
dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); |
359 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
360 |
|
|
361 |
if (dc_size > 8) |
if (dc_size > 8) { |
|
{ |
|
362 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); // marker |
363 |
} |
} |
364 |
|
|
365 |
block[i*64 + 0] = dc_dif; |
block[i*64 + 0] = dc_dif; |
366 |
start_coeff = 1; |
start_coeff = 1; |
367 |
} |
} else { |
|
else |
|
|
{ |
|
368 |
start_coeff = 0; |
start_coeff = 0; |
369 |
} |
} |
370 |
|
|
371 |
start_timer(); |
start_timer(); |
372 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
373 |
{ |
{ |
374 |
get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
375 |
|
start_coeff); |
376 |
} |
} |
377 |
stop_coding_timer(); |
stop_coding_timer(); |
378 |
|
|
381 |
stop_prediction_timer(); |
stop_prediction_timer(); |
382 |
|
|
383 |
start_timer(); |
start_timer(); |
384 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
385 |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
386 |
} |
} else { |
|
else |
|
|
{ |
|
387 |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
388 |
} |
} |
389 |
stop_iquant_timer(); |
stop_iquant_timer(); |
393 |
stop_idct_timer(); |
stop_idct_timer(); |
394 |
} |
} |
395 |
|
|
396 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
397 |
next_block = stride; |
next_block = stride; |
398 |
stride *= 2; |
stride *= 2; |
399 |
} |
} |
420 |
|
|
421 |
// decode an inter macroblock |
// decode an inter macroblock |
422 |
|
|
423 |
void decoder_mbinter(DECODER * dec, |
void |
424 |
|
decoder_mbinter(DECODER * dec, |
425 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
426 |
const uint32_t x_pos, |
const uint32_t x_pos, |
427 |
const uint32_t y_pos, |
const uint32_t y_pos, |
447 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
448 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
449 |
|
|
450 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
|
{ |
|
451 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
452 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
453 |
|
|
454 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
455 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
456 |
} |
} else { |
|
else |
|
|
{ |
|
457 |
int sum; |
int sum; |
458 |
|
|
459 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
460 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = |
461 |
|
(sum == |
462 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
463 |
|
(ABS(sum) / 16) * 2)); |
464 |
|
|
465 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
466 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = |
467 |
|
(sum == |
468 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
469 |
|
(ABS(sum) / 16) * 2)); |
470 |
} |
} |
471 |
|
|
472 |
start_timer(); |
start_timer(); |
473 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, 16 * y_pos, |
474 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
475 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, |
476 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, |
477 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
rounding); |
478 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, |
479 |
|
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
480 |
|
rounding); |
481 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, |
482 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
483 |
|
rounding); |
484 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
485 |
|
uv_dx, uv_dy, stride2, rounding); |
486 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
487 |
|
uv_dx, uv_dy, stride2, rounding); |
488 |
stop_comp_timer(); |
stop_comp_timer(); |
489 |
|
|
490 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
491 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
492 |
{ |
{ |
493 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
497 |
stop_coding_timer(); |
stop_coding_timer(); |
498 |
|
|
499 |
start_timer(); |
start_timer(); |
500 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
501 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
502 |
} |
} else { |
|
else |
|
|
{ |
|
503 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
504 |
} |
} |
505 |
stop_iquant_timer(); |
stop_iquant_timer(); |
510 |
} |
} |
511 |
} |
} |
512 |
|
|
513 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
514 |
next_block = stride; |
next_block = stride; |
515 |
stride *= 2; |
stride *= 2; |
516 |
} |
} |
532 |
} |
} |
533 |
|
|
534 |
|
|
535 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
void |
536 |
|
decoder_iframe(DECODER * dec, |
537 |
|
Bitstream * bs, |
538 |
|
int quant, |
539 |
|
int intra_dc_threshold) |
540 |
{ |
{ |
541 |
|
|
542 |
uint32_t x, y; |
uint32_t x, y; |
543 |
|
|
544 |
for (y = 0; y < dec->mb_height; y++) |
for (y = 0; y < dec->mb_height; y++) { |
545 |
{ |
for (x = 0; x < dec->mb_width; x++) { |
|
for (x = 0; x < dec->mb_width; x++) |
|
|
{ |
|
546 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
547 |
|
|
548 |
uint32_t mcbpc; |
uint32_t mcbpc; |
557 |
|
|
558 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
559 |
|
|
560 |
if (mb->mode == MODE_STUFFING) |
if (mb->mode == MODE_STUFFING) { |
|
{ |
|
561 |
DEBUG("-- STUFFING ?"); |
DEBUG("-- STUFFING ?"); |
562 |
continue; |
continue; |
563 |
} |
} |
565 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
566 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
567 |
|
|
568 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
569 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
570 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
571 |
quant = 31; |
quant = 31; |
572 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
573 |
quant = 1; |
quant = 1; |
574 |
} |
} |
575 |
} |
} |
576 |
mb->quant = quant; |
mb->quant = quant; |
577 |
|
|
578 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
579 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
580 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DEBUG1("deci: field_dct: ", mb->field_dct); |
581 |
} |
} |
582 |
|
|
583 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
584 |
|
intra_dc_threshold); |
585 |
} |
} |
586 |
} |
} |
587 |
|
|
588 |
} |
} |
589 |
|
|
590 |
|
|
591 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
void |
592 |
|
get_motion_vector(DECODER * dec, |
593 |
|
Bitstream * bs, |
594 |
|
int x, |
595 |
|
int y, |
596 |
|
int k, |
597 |
|
VECTOR * mv, |
598 |
|
int fcode) |
599 |
{ |
{ |
600 |
|
|
601 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
604 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
605 |
|
|
606 |
VECTOR pmv[4]; |
VECTOR pmv[4]; |
607 |
uint32_t psad[4]; |
int32_t psad[4]; |
608 |
|
|
609 |
int mv_x, mv_y; |
int mv_x, mv_y; |
610 |
int pmv_x, pmv_y; |
int pmv_x, pmv_y; |
621 |
mv_x += pmv_x; |
mv_x += pmv_x; |
622 |
mv_y += pmv_y; |
mv_y += pmv_y; |
623 |
|
|
624 |
if (mv_x < low) |
if (mv_x < low) { |
|
{ |
|
625 |
mv_x += range; |
mv_x += range; |
626 |
} |
} else if (mv_x > high) { |
|
else if (mv_x > high) |
|
|
{ |
|
627 |
mv_x -= range; |
mv_x -= range; |
628 |
} |
} |
629 |
|
|
630 |
if (mv_y < low) |
if (mv_y < low) { |
|
{ |
|
631 |
mv_y += range; |
mv_y += range; |
632 |
} |
} else if (mv_y > high) { |
|
else if (mv_y > high) |
|
|
{ |
|
633 |
mv_y -= range; |
mv_y -= range; |
634 |
} |
} |
635 |
|
|
639 |
} |
} |
640 |
|
|
641 |
|
|
642 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
void |
643 |
|
decoder_pframe(DECODER * dec, |
644 |
|
Bitstream * bs, |
645 |
|
int rounding, |
646 |
|
int quant, |
647 |
|
int fcode, |
648 |
|
int intra_dc_threshold) |
649 |
{ |
{ |
650 |
|
|
651 |
uint32_t x, y; |
uint32_t x, y; |
652 |
|
|
653 |
start_timer(); |
start_timer(); |
654 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
655 |
|
dec->width, dec->height, dec->interlacing); |
656 |
stop_edges_timer(); |
stop_edges_timer(); |
657 |
|
|
658 |
for (y = 0; y < dec->mb_height; y++) |
for (y = 0; y < dec->mb_height; y++) { |
659 |
{ |
for (x = 0; x < dec->mb_width; x++) { |
|
for (x = 0; x < dec->mb_width; x++) |
|
|
{ |
|
660 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
661 |
|
|
662 |
if (!BitstreamGetBit(bs)) // not_coded |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
663 |
|
if (!(BitstreamGetBit(bs))) // not_coded |
664 |
{ |
{ |
665 |
uint32_t mcbpc; |
uint32_t mcbpc; |
666 |
uint32_t cbpc; |
uint32_t cbpc; |
676 |
|
|
677 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
678 |
|
|
679 |
if (intra) |
if (intra) { |
|
{ |
|
680 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
681 |
} |
} |
682 |
|
|
683 |
if (mb->mode == MODE_STUFFING) |
if (mb->mode == MODE_STUFFING) { |
|
{ |
|
684 |
DEBUG("-- STUFFING ?"); |
DEBUG("-- STUFFING ?"); |
685 |
continue; |
continue; |
686 |
} |
} |
688 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
689 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
690 |
|
|
691 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
|
{ |
|
692 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
693 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
694 |
quant = 31; |
quant = 31; |
695 |
} |
} else if (mb->quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
696 |
quant = 1; |
quant = 1; |
697 |
} |
} |
698 |
} |
} |
699 |
mb->quant = quant; |
mb->quant = quant; |
700 |
|
|
701 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
702 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
703 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DEBUG1("decp: field_dct: ", mb->field_dct); |
704 |
|
|
705 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
706 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
707 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DEBUG1("decp: field_pred: ", mb->field_pred); |
708 |
|
|
709 |
if (mb->field_pred) |
if (mb->field_pred) { |
|
{ |
|
710 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
711 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
712 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
715 |
} |
} |
716 |
} |
} |
717 |
|
|
718 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
719 |
{ |
if (dec->interlacing && mb->field_pred) { |
720 |
if (dec->interlacing && mb->field_pred) |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
721 |
{ |
fcode); |
722 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
723 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); |
fcode); |
724 |
|
} else { |
725 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
726 |
|
fcode); |
727 |
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
728 |
|
mb->mvs[0].x; |
729 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
730 |
|
mb->mvs[0].y; |
731 |
} |
} |
732 |
else |
} else if (mb->mode == |
733 |
{ |
MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */ ) { |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
|
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
|
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
|
} |
|
|
} |
|
|
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
|
|
{ |
|
734 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
735 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
736 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
737 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
738 |
} |
} else // MODE_INTRA, MODE_INTRA_Q |
|
else // MODE_INTRA, MODE_INTRA_Q |
|
739 |
{ |
{ |
740 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
741 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
0; |
742 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
743 |
|
0; |
744 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
745 |
|
intra_dc_threshold); |
746 |
continue; |
continue; |
747 |
} |
} |
748 |
|
|
749 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
750 |
} |
rounding); |
751 |
else // not coded |
} else // not coded |
752 |
{ |
{ |
753 |
|
//DEBUG2("P-frame MB at (X,Y)=",x,y); |
754 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
755 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
756 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
759 |
|
|
760 |
start_timer(); |
start_timer(); |
761 |
|
|
762 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
763 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
(16 * x), |
764 |
dec->edged_width); |
dec->refn[0].y + (16 * y) * dec->edged_width + |
765 |
|
(16 * x), dec->edged_width); |
766 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
|
767 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x+8), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
768 |
dec->edged_width); |
(16 * x + 8), |
769 |
|
dec->refn[0].y + (16 * y) * dec->edged_width + |
770 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
(16 * x + 8), dec->edged_width); |
771 |
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x), |
|
772 |
dec->edged_width); |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
773 |
|
(16 * x), |
774 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
dec->refn[0].y + (16 * y + |
775 |
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x+8), |
8) * dec->edged_width + |
776 |
dec->edged_width); |
(16 * x), dec->edged_width); |
777 |
|
|
778 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
779 |
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
(16 * x + 8), |
780 |
|
dec->refn[0].y + (16 * y + |
781 |
|
8) * dec->edged_width + |
782 |
|
(16 * x + 8), dec->edged_width); |
783 |
|
|
784 |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
785 |
|
(8 * x), |
786 |
|
dec->refn[0].u + |
787 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
788 |
dec->edged_width/2); |
dec->edged_width/2); |
789 |
|
|
790 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
791 |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
(8 * x), |
792 |
|
dec->refn[0].v + |
793 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
794 |
dec->edged_width/2); |
dec->edged_width/2); |
795 |
|
|
796 |
stop_transfer_timer(); |
stop_transfer_timer(); |
799 |
} |
} |
800 |
} |
} |
801 |
|
|
802 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
|
803 |
|
// add by MinChen <chenm001@163.com> |
804 |
|
// decode B-frame motion vector |
805 |
|
void |
806 |
|
get_b_motion_vector(DECODER * dec, |
807 |
|
Bitstream * bs, |
808 |
|
int x, |
809 |
|
int y, |
810 |
|
VECTOR * mv, |
811 |
|
int fcode, |
812 |
|
const VECTOR pmv) |
813 |
|
{ |
814 |
|
int scale_fac = 1 << (fcode - 1); |
815 |
|
int high = (32 * scale_fac) - 1; |
816 |
|
int low = ((-32) * scale_fac); |
817 |
|
int range = (64 * scale_fac); |
818 |
|
|
819 |
|
int mv_x, mv_y; |
820 |
|
int pmv_x, pmv_y; |
821 |
|
|
822 |
|
pmv_x = pmv.x; |
823 |
|
pmv_y = pmv.y; |
824 |
|
|
825 |
|
mv_x = get_mv(bs, fcode); |
826 |
|
mv_y = get_mv(bs, fcode); |
827 |
|
|
828 |
|
mv_x += pmv_x; |
829 |
|
mv_y += pmv_y; |
830 |
|
|
831 |
|
if (mv_x < low) { |
832 |
|
mv_x += range; |
833 |
|
} else if (mv_x > high) { |
834 |
|
mv_x -= range; |
835 |
|
} |
836 |
|
|
837 |
|
if (mv_y < low) { |
838 |
|
mv_y += range; |
839 |
|
} else if (mv_y > high) { |
840 |
|
mv_y -= range; |
841 |
|
} |
842 |
|
|
843 |
|
mv->x = mv_x; |
844 |
|
mv->y = mv_y; |
845 |
|
} |
846 |
|
|
847 |
|
|
848 |
|
// add by MinChen <chenm001@163.com> |
849 |
|
// decode an B-frame forward & backward inter macroblock |
850 |
|
void |
851 |
|
decoder_bf_mbinter(DECODER * dec, |
852 |
|
const MACROBLOCK * pMB, |
853 |
|
const uint32_t x_pos, |
854 |
|
const uint32_t y_pos, |
855 |
|
const uint32_t cbp, |
856 |
|
Bitstream * bs, |
857 |
|
const uint32_t quant, |
858 |
|
const uint8_t ref) |
859 |
|
{ |
860 |
|
|
861 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
862 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
863 |
|
|
864 |
|
uint32_t stride = dec->edged_width; |
865 |
|
uint32_t stride2 = stride / 2; |
866 |
|
uint32_t next_block = stride * 8; |
867 |
|
uint32_t i; |
868 |
|
uint32_t iQuant = pMB->quant; |
869 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
870 |
|
int uv_dx, uv_dy; |
871 |
|
|
872 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
873 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
874 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
875 |
|
|
876 |
|
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
877 |
|
uv_dx = pMB->mvs[0].x; |
878 |
|
uv_dy = pMB->mvs[0].y; |
879 |
|
|
880 |
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
881 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
882 |
|
} else { |
883 |
|
int sum; |
884 |
|
|
885 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
886 |
|
uv_dx = |
887 |
|
(sum == |
888 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
889 |
|
(ABS(sum) / 16) * 2)); |
890 |
|
|
891 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
892 |
|
uv_dy = |
893 |
|
(sum == |
894 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
895 |
|
(ABS(sum) / 16) * 2)); |
896 |
|
} |
897 |
|
|
898 |
|
start_timer(); |
899 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos, |
900 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
901 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
902 |
|
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
903 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
904 |
|
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
905 |
|
0); |
906 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
907 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
908 |
|
0); |
909 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
910 |
|
uv_dx, uv_dy, stride2, 0); |
911 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
912 |
|
uv_dx, uv_dy, stride2, 0); |
913 |
|
stop_comp_timer(); |
914 |
|
|
915 |
|
for (i = 0; i < 6; i++) { |
916 |
|
if (cbp & (1 << (5 - i))) // coded |
917 |
|
{ |
918 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
919 |
|
|
920 |
|
start_timer(); |
921 |
|
get_inter_block(bs, &block[i * 64]); |
922 |
|
stop_coding_timer(); |
923 |
|
|
924 |
|
start_timer(); |
925 |
|
if (dec->quant_type == 0) { |
926 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
927 |
|
} else { |
928 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
929 |
|
} |
930 |
|
stop_iquant_timer(); |
931 |
|
|
932 |
|
start_timer(); |
933 |
|
idct(&data[i * 64]); |
934 |
|
stop_idct_timer(); |
935 |
|
} |
936 |
|
} |
937 |
|
|
938 |
|
if (dec->interlacing && pMB->field_dct) { |
939 |
|
next_block = stride; |
940 |
|
stride *= 2; |
941 |
|
} |
942 |
|
|
943 |
|
start_timer(); |
944 |
|
if (cbp & 32) |
945 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
946 |
|
if (cbp & 16) |
947 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
948 |
|
if (cbp & 8) |
949 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
950 |
|
if (cbp & 4) |
951 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
952 |
|
if (cbp & 2) |
953 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
954 |
|
if (cbp & 1) |
955 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
956 |
|
stop_transfer_timer(); |
957 |
|
} |
958 |
|
|
959 |
|
|
960 |
|
// add by MinChen <chenm001@163.com> |
961 |
|
// decode an B-frame direct & inter macroblock |
962 |
|
void |
963 |
|
decoder_bf_interpolate_mbinter(DECODER * dec, |
964 |
|
IMAGE forward, |
965 |
|
IMAGE backward, |
966 |
|
const MACROBLOCK * pMB, |
967 |
|
const uint32_t x_pos, |
968 |
|
const uint32_t y_pos, |
969 |
|
const uint32_t cbp, |
970 |
|
Bitstream * bs) |
971 |
|
{ |
972 |
|
|
973 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
974 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
975 |
|
|
976 |
|
uint32_t stride = dec->edged_width; |
977 |
|
uint32_t stride2 = stride / 2; |
978 |
|
uint32_t next_block = stride * 8; |
979 |
|
uint32_t iQuant = pMB->quant; |
980 |
|
int uv_dx, uv_dy; |
981 |
|
int b_uv_dx, b_uv_dy; |
982 |
|
uint32_t i; |
983 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
984 |
|
|
985 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
986 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
987 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
988 |
|
|
989 |
|
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
990 |
|
uv_dx = pMB->mvs[0].x; |
991 |
|
uv_dy = pMB->mvs[0].y; |
992 |
|
|
993 |
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
994 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
995 |
|
|
996 |
|
b_uv_dx = pMB->b_mvs[0].x; |
997 |
|
b_uv_dy = pMB->b_mvs[0].y; |
998 |
|
|
999 |
|
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
1000 |
|
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
1001 |
|
} else { |
1002 |
|
int sum; |
1003 |
|
|
1004 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1005 |
|
uv_dx = |
1006 |
|
(sum == |
1007 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1008 |
|
(ABS(sum) / 16) * 2)); |
1009 |
|
|
1010 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1011 |
|
uv_dy = |
1012 |
|
(sum == |
1013 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1014 |
|
(ABS(sum) / 16) * 2)); |
1015 |
|
|
1016 |
|
sum = |
1017 |
|
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
1018 |
|
pMB->b_mvs[3].x; |
1019 |
|
b_uv_dx = |
1020 |
|
(sum == |
1021 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1022 |
|
(ABS(sum) / 16) * 2)); |
1023 |
|
|
1024 |
|
sum = |
1025 |
|
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
1026 |
|
pMB->b_mvs[3].y; |
1027 |
|
b_uv_dy = |
1028 |
|
(sum == |
1029 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1030 |
|
(ABS(sum) / 16) * 2)); |
1031 |
|
} |
1032 |
|
|
1033 |
|
|
1034 |
|
start_timer(); |
1035 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1036 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1037 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1038 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1039 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1040 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1041 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1042 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1043 |
|
0); |
1044 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1045 |
|
uv_dy, stride2, 0); |
1046 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1047 |
|
uv_dy, stride2, 0); |
1048 |
|
|
1049 |
|
|
1050 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
1051 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1052 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1053 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1054 |
|
0); |
1055 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, |
1056 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1057 |
|
stride, 0); |
1058 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1059 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1060 |
|
stride, 0); |
1061 |
|
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
1062 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1063 |
|
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1064 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1065 |
|
|
1066 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, |
1067 |
|
stride); |
1068 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
1069 |
|
stride); |
1070 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
1071 |
|
stride); |
1072 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
1073 |
|
16 * y_pos + 8, stride); |
1074 |
|
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
1075 |
|
stride2); |
1076 |
|
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, |
1077 |
|
stride2); |
1078 |
|
|
1079 |
|
stop_comp_timer(); |
1080 |
|
|
1081 |
|
for (i = 0; i < 6; i++) { |
1082 |
|
if (cbp & (1 << (5 - i))) // coded |
1083 |
|
{ |
1084 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1085 |
|
|
1086 |
|
start_timer(); |
1087 |
|
get_inter_block(bs, &block[i * 64]); |
1088 |
|
stop_coding_timer(); |
1089 |
|
|
1090 |
|
start_timer(); |
1091 |
|
if (dec->quant_type == 0) { |
1092 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1093 |
|
} else { |
1094 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1095 |
|
} |
1096 |
|
stop_iquant_timer(); |
1097 |
|
|
1098 |
|
start_timer(); |
1099 |
|
idct(&data[i * 64]); |
1100 |
|
stop_idct_timer(); |
1101 |
|
} |
1102 |
|
} |
1103 |
|
|
1104 |
|
if (dec->interlacing && pMB->field_dct) { |
1105 |
|
next_block = stride; |
1106 |
|
stride *= 2; |
1107 |
|
} |
1108 |
|
|
1109 |
|
start_timer(); |
1110 |
|
if (cbp & 32) |
1111 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1112 |
|
if (cbp & 16) |
1113 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1114 |
|
if (cbp & 8) |
1115 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1116 |
|
if (cbp & 4) |
1117 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1118 |
|
if (cbp & 2) |
1119 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1120 |
|
if (cbp & 1) |
1121 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1122 |
|
stop_transfer_timer(); |
1123 |
|
} |
1124 |
|
|
1125 |
|
|
1126 |
|
// add by MinChen <chenm001@163.com> |
1127 |
|
// for decode B-frame dbquant |
1128 |
|
int32_t __inline |
1129 |
|
get_dbquant(Bitstream * bs) |
1130 |
|
{ |
1131 |
|
if (!BitstreamGetBit(bs)) // '0' |
1132 |
|
return (0); |
1133 |
|
else if (!BitstreamGetBit(bs)) // '10' |
1134 |
|
return (-2); |
1135 |
|
else |
1136 |
|
return (2); // '11' |
1137 |
|
} |
1138 |
|
|
1139 |
|
// add by MinChen <chenm001@163.com> |
1140 |
|
// for decode B-frame mb_type |
1141 |
|
// bit ret_value |
1142 |
|
// 1 0 |
1143 |
|
// 01 1 |
1144 |
|
// 001 2 |
1145 |
|
// 0001 3 |
1146 |
|
int32_t __inline |
1147 |
|
get_mbtype(Bitstream * bs) |
1148 |
|
{ |
1149 |
|
int32_t mb_type; |
1150 |
|
|
1151 |
|
for (mb_type = 0; mb_type <= 3; mb_type++) { |
1152 |
|
if (BitstreamGetBit(bs)) |
1153 |
|
break; |
1154 |
|
} |
1155 |
|
|
1156 |
|
if (mb_type <= 3) |
1157 |
|
return (mb_type); |
1158 |
|
else |
1159 |
|
return (-1); |
1160 |
|
} |
1161 |
|
|
1162 |
|
void |
1163 |
|
decoder_bframe(DECODER * dec, |
1164 |
|
Bitstream * bs, |
1165 |
|
int quant, |
1166 |
|
int fcode_forward, |
1167 |
|
int fcode_backward) |
1168 |
|
{ |
1169 |
|
|
1170 |
|
uint32_t x, y; |
1171 |
|
VECTOR mv, zeromv; |
1172 |
|
|
1173 |
|
start_timer(); |
1174 |
|
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1175 |
|
dec->width, dec->height, dec->interlacing); |
1176 |
|
//image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
1177 |
|
stop_edges_timer(); |
1178 |
|
|
1179 |
|
|
1180 |
|
for (y = 0; y < dec->mb_height; y++) { |
1181 |
|
// Initialize Pred Motion Vector |
1182 |
|
dec->p_fmv.x = dec->p_fmv.y = dec->p_bmv.x = dec->p_bmv.y = 0; |
1183 |
|
for (x = 0; x < dec->mb_width; x++) { |
1184 |
|
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1185 |
|
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1186 |
|
|
1187 |
|
mb->mvs[0].x = mb->mvs[0].y = zeromv.x = zeromv.y = mv.x = mv.y = |
1188 |
|
0; |
1189 |
|
|
1190 |
|
// the last P_VOP is skip macroblock ? |
1191 |
|
if (last_mb->mode == MODE_NOT_CODED) { |
1192 |
|
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
1193 |
|
mb->mb_type = MODE_FORWARD; |
1194 |
|
mb->cbp = 0; |
1195 |
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1196 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1197 |
|
mb->quant = 8; |
1198 |
|
|
1199 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1200 |
|
continue; |
1201 |
|
} |
1202 |
|
//t=BitstreamShowBits(bs,32); |
1203 |
|
|
1204 |
|
if (!BitstreamGetBit(bs)) { // modb=='0' |
1205 |
|
const uint8_t modb2 = BitstreamGetBit(bs); |
1206 |
|
|
1207 |
|
mb->mb_type = get_mbtype(bs); |
1208 |
|
|
1209 |
|
if (!modb2) { // modb=='00' |
1210 |
|
mb->cbp = BitstreamGetBits(bs, 6); |
1211 |
|
} else { |
1212 |
|
mb->cbp = 0; |
1213 |
|
} |
1214 |
|
if (mb->mb_type && mb->cbp) { |
1215 |
|
quant += get_dbquant(bs); |
1216 |
|
|
1217 |
|
if (quant > 31) { |
1218 |
|
quant = 31; |
1219 |
|
} else if (mb->quant < 1) { |
1220 |
|
quant = 1; |
1221 |
|
} |
1222 |
|
} else { |
1223 |
|
quant = 8; |
1224 |
|
} |
1225 |
|
mb->quant = quant; |
1226 |
|
} else { |
1227 |
|
mb->mb_type = MODE_DIRECT_NONE_MV; |
1228 |
|
mb->cbp = 0; |
1229 |
|
} |
1230 |
|
|
1231 |
|
mb->mode = MODE_INTER; |
1232 |
|
//DEBUG1("Switch bm_type=",mb->mb_type); |
1233 |
|
|
1234 |
|
switch (mb->mb_type) { |
1235 |
|
case MODE_DIRECT: |
1236 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], 1, zeromv); |
1237 |
|
|
1238 |
|
case MODE_DIRECT_NONE_MV: |
1239 |
|
{ // Because this file is a C file not C++ so I use '{' to define var |
1240 |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = |
1241 |
|
dec->time_pp; |
1242 |
|
int i; |
1243 |
|
|
1244 |
|
for (i = 0; i < 4; i++) { |
1245 |
|
mb->mvs[i].x = |
1246 |
|
(int32_t) ((TRB * last_mb->mvs[i].x) / TRD + |
1247 |
|
mb->mvs[0].x); |
1248 |
|
mb->b_mvs[i].x = |
1249 |
|
(int32_t) ((mb->mvs[0].x == |
1250 |
|
0) ? ((TRB - |
1251 |
|
TRD) * last_mb->mvs[i].x) / |
1252 |
|
TRD : mb->mvs[i].x - last_mb->mvs[i].x); |
1253 |
|
mb->mvs[i].y = |
1254 |
|
(int32_t) ((TRB * last_mb->mvs[i].y) / TRD + |
1255 |
|
mb->mvs[0].y); |
1256 |
|
mb->b_mvs[i].y = |
1257 |
|
(int32_t) ((mb->mvs[0].y == |
1258 |
|
0) ? ((TRB - |
1259 |
|
TRD) * last_mb->mvs[i].y) / |
1260 |
|
TRD : mb->mvs[i].y - last_mb->mvs[i].y); |
1261 |
|
} |
1262 |
|
//DEBUG("B-frame Direct!\n"); |
1263 |
|
} |
1264 |
|
mb->mode = MODE_INTER4V; |
1265 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1266 |
|
mb, x, y, mb->cbp, bs); |
1267 |
|
break; |
1268 |
|
|
1269 |
|
case MODE_INTERPOLATE: |
1270 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1271 |
|
dec->p_fmv); |
1272 |
|
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
1273 |
|
mb->mvs[0].x; |
1274 |
|
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
1275 |
|
mb->mvs[0].y; |
1276 |
|
|
1277 |
|
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1278 |
|
fcode_backward, dec->p_bmv); |
1279 |
|
dec->p_bmv.x = mb->b_mvs[1].x = mb->b_mvs[2].x = |
1280 |
|
mb->b_mvs[3].x = mb->b_mvs[0].x; |
1281 |
|
dec->p_bmv.y = mb->b_mvs[1].y = mb->b_mvs[2].y = |
1282 |
|
mb->b_mvs[3].y = mb->b_mvs[0].y; |
1283 |
|
|
1284 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1285 |
|
mb, x, y, mb->cbp, bs); |
1286 |
|
//DEBUG("B-frame Bidir!\n"); |
1287 |
|
break; |
1288 |
|
|
1289 |
|
case MODE_BACKWARD: |
1290 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1291 |
|
dec->p_bmv); |
1292 |
|
dec->p_bmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
1293 |
|
mb->mvs[0].x; |
1294 |
|
dec->p_bmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
1295 |
|
mb->mvs[0].y; |
1296 |
|
|
1297 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1298 |
|
//DEBUG("B-frame Backward!\n"); |
1299 |
|
break; |
1300 |
|
|
1301 |
|
case MODE_FORWARD: |
1302 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1303 |
|
dec->p_fmv); |
1304 |
|
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
1305 |
|
mb->mvs[0].x; |
1306 |
|
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
1307 |
|
mb->mvs[0].y; |
1308 |
|
|
1309 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1310 |
|
//DEBUG("B-frame Forward!\n"); |
1311 |
|
break; |
1312 |
|
|
1313 |
|
default: |
1314 |
|
DEBUG1("Not support B-frame mb_type =", mb->mb_type); |
1315 |
|
} |
1316 |
|
|
1317 |
|
} // end of FOR |
1318 |
|
} |
1319 |
|
} |
1320 |
|
|
1321 |
|
// swap two MACROBLOCK array |
1322 |
|
void |
1323 |
|
mb_swap(MACROBLOCK ** mb1, |
1324 |
|
MACROBLOCK ** mb2) |
1325 |
|
{ |
1326 |
|
MACROBLOCK *temp = *mb1; |
1327 |
|
|
1328 |
|
*mb1 = *mb2; |
1329 |
|
*mb2 = temp; |
1330 |
|
} |
1331 |
|
|
1332 |
|
int |
1333 |
|
decoder_decode(DECODER * dec, |
1334 |
|
XVID_DEC_FRAME * frame) |
1335 |
{ |
{ |
1336 |
|
|
1337 |
Bitstream bs; |
Bitstream bs; |
1338 |
uint32_t rounding; |
uint32_t rounding; |
1339 |
uint32_t quant; |
uint32_t quant; |
1340 |
uint32_t fcode; |
uint32_t fcode_forward; |
1341 |
|
uint32_t fcode_backward; |
1342 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1343 |
uint32_t vop_type; |
uint32_t vop_type; |
1344 |
|
|
1348 |
|
|
1349 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
1350 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
1351 |
vop_type=BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold); |
dec->frames++; |
1352 |
|
vop_type = |
1353 |
|
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
1354 |
|
&fcode_backward, &intra_dc_threshold); |
1355 |
|
|
1356 |
if (vop_type==I_VOP || vop_type==P_VOP){ |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
|
|
image_swap(&dec->cur, &dec->refn[0]); |
|
|
} |
|
1357 |
|
|
1358 |
switch (vop_type) |
switch (vop_type) { |
|
{ |
|
1359 |
case P_VOP : |
case P_VOP : |
1360 |
decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1361 |
|
intra_dc_threshold); |
1362 |
|
printf("P_VOP Time=%ld\n",dec->time); |
1363 |
|
DEBUG1("P_VOP Time=", dec->time); |
1364 |
break; |
break; |
1365 |
|
|
1366 |
case I_VOP : |
case I_VOP : |
|
//DEBUG1("",intra_dc_threshold); |
|
1367 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1368 |
|
printf("I_VOP Time=%ld\n",dec->time); |
1369 |
|
DEBUG1("I_VOP Time=", dec->time); |
1370 |
break; |
break; |
1371 |
|
|
1372 |
case B_VOP : // ignore |
case B_VOP: |
1373 |
|
#ifdef BFRAMES_DEC |
1374 |
|
if (dec->time_pp > dec->time_bp) { |
1375 |
|
printf("I_VOP Time=%ld\n",dec->time); |
1376 |
|
DEBUG1("B_VOP Time=", dec->time); |
1377 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1378 |
|
} else { |
1379 |
|
DEBUG("broken B-frame!"); |
1380 |
|
} |
1381 |
|
#else |
1382 |
|
printf("Che minchia ne so\n"); |
1383 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1384 |
|
#endif |
1385 |
break; |
break; |
1386 |
|
|
1387 |
case N_VOP : // vop not coded |
case N_VOP : // vop not coded |
1388 |
|
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
1389 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1390 |
|
printf("N_VOP vop not coded\n"); |
1391 |
break; |
break; |
1392 |
|
|
1393 |
default : |
default : |
1396 |
|
|
1397 |
frame->length = BitstreamPos(&bs) / 8; |
frame->length = BitstreamPos(&bs) / 8; |
1398 |
|
|
1399 |
start_timer(); |
#ifdef BFRAMES_DEC |
1400 |
|
// test if no B_VOP |
1401 |
|
if (dec->low_delay) { |
1402 |
|
#endif |
1403 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1404 |
frame->image, frame->stride, frame->colorspace); |
frame->image, frame->stride, frame->colorspace); |
1405 |
|
#ifdef BFRAMES_DEC |
1406 |
|
} else { |
1407 |
|
if (dec->frames >= 1) { |
1408 |
|
start_timer(); |
1409 |
|
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
1410 |
|
image_output(&dec->refn[0], dec->width, dec->height, |
1411 |
|
dec->edged_width, frame->image, frame->stride, |
1412 |
|
frame->colorspace); |
1413 |
|
} else if (vop_type == B_VOP) { |
1414 |
|
image_output(&dec->cur, dec->width, dec->height, |
1415 |
|
dec->edged_width, frame->image, frame->stride, |
1416 |
|
frame->colorspace); |
1417 |
|
} |
1418 |
stop_conv_timer(); |
stop_conv_timer(); |
1419 |
|
} |
1420 |
|
} |
1421 |
|
#endif |
1422 |
|
|
1423 |
|
if (vop_type == I_VOP || vop_type == P_VOP) { |
1424 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1425 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1426 |
|
// swap MACROBLOCK |
1427 |
|
if (dec->low_delay && vop_type == P_VOP) |
1428 |
|
mb_swap(&dec->mbs, &dec->last_mbs); |
1429 |
|
} |
1430 |
|
|
1431 |
emms(); |
emms(); |
1432 |
|
|
1433 |
stop_global_timer(); |
stop_global_timer(); |
1434 |
|
|
1435 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
|
|
|
1436 |
} |
} |