32 |
* |
* |
33 |
* History: |
* History: |
34 |
* |
* |
35 |
|
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
36 |
|
* MinChen <chenm001@163.com> |
37 |
|
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
38 |
|
* Fix a little bug for low_delay flage |
39 |
|
* MinChen <chenm001@163.com> |
40 |
|
* 28.06.2002 added basic resync support to iframe/pframe_decode() |
41 |
|
* 22.06.2002 added primative N_VOP support |
42 |
|
* #define BFRAMES_DEC now enables Minchen's bframe decoder |
43 |
* 08.05.2002 add low_delay support for B_VOP decode |
* 08.05.2002 add low_delay support for B_VOP decode |
44 |
* MinChen <chenm001@163.com> |
* MinChen <chenm001@163.com> |
45 |
* 05.05.2002 fix some B-frame decode problem |
* 05.05.2002 fix some B-frame decode problem |
62 |
#include <stdlib.h> |
#include <stdlib.h> |
63 |
#include <string.h> |
#include <string.h> |
64 |
|
|
65 |
|
#ifdef BFRAMES_DEC_DEBUG |
66 |
|
#define BFRAMES_DEC |
67 |
|
#endif |
68 |
|
|
69 |
#include "xvid.h" |
#include "xvid.h" |
70 |
#include "portab.h" |
#include "portab.h" |
71 |
|
|
89 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
90 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
91 |
|
|
92 |
int decoder_create(XVID_DEC_PARAM * param) |
int |
93 |
|
decoder_create(XVID_DEC_PARAM * param) |
94 |
{ |
{ |
95 |
DECODER * dec; |
DECODER * dec; |
96 |
|
|
97 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
98 |
if (dec == NULL) |
if (dec == NULL) { |
|
{ |
|
99 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
100 |
} |
} |
101 |
param->handle = dec; |
param->handle = dec; |
108 |
|
|
109 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
110 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
111 |
|
dec->low_delay = 0; |
112 |
|
|
113 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
|
{ |
|
114 |
xvid_free(dec); |
xvid_free(dec); |
115 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
116 |
} |
} |
117 |
|
|
118 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
119 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
120 |
xvid_free(dec); |
xvid_free(dec); |
121 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
122 |
} |
} |
123 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
124 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
125 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
|
{ |
|
126 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
127 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
128 |
xvid_free(dec); |
xvid_free(dec); |
129 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
130 |
} |
} |
131 |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
|
{ |
|
132 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
133 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
134 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
136 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
137 |
} |
} |
138 |
|
|
139 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
if (image_create(&dec->refh, dec->edged_width, dec->edged_height)) { |
140 |
if (dec->mbs == NULL) |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
141 |
{ |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
142 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
143 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
144 |
|
xvid_free(dec); |
145 |
|
return XVID_ERR_MEMORY; |
146 |
|
} |
147 |
|
|
148 |
|
dec->mbs = |
149 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
150 |
|
CACHE_LINE); |
151 |
|
if (dec->mbs == NULL) { |
152 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
153 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
154 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
155 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
156 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
157 |
xvid_free(dec); |
xvid_free(dec); |
158 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
159 |
} |
} |
160 |
|
|
161 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
162 |
|
|
163 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
164 |
// for skip MB flag |
// for skip MB flag |
165 |
dec->last_mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
dec->last_mbs = |
166 |
if (dec->last_mbs == NULL) |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
167 |
{ |
CACHE_LINE); |
168 |
|
if (dec->last_mbs == NULL) { |
169 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
170 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
171 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
172 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
173 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
174 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
175 |
xvid_free(dec); |
xvid_free(dec); |
176 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
177 |
} |
} |
178 |
|
|
179 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
180 |
|
|
181 |
init_timer(); |
init_timer(); |
182 |
|
|
183 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
189 |
} |
} |
190 |
|
|
191 |
|
|
192 |
int decoder_destroy(DECODER * dec) |
int |
193 |
|
decoder_destroy(DECODER * dec) |
194 |
{ |
{ |
195 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
196 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
197 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
198 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
199 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
200 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
201 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
202 |
xvid_free(dec); |
xvid_free(dec); |
203 |
|
|
207 |
|
|
208 |
|
|
209 |
|
|
210 |
static const int32_t dquant_table[4] = |
static const int32_t dquant_table[4] = { |
|
{ |
|
211 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
212 |
}; |
}; |
213 |
|
|
216 |
|
|
217 |
// decode an intra macroblock |
// decode an intra macroblock |
218 |
|
|
219 |
void decoder_mbintra(DECODER * dec, |
void |
220 |
|
decoder_mbintra(DECODER * dec, |
221 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
222 |
const uint32_t x_pos, |
const uint32_t x_pos, |
223 |
const uint32_t y_pos, |
const uint32_t y_pos, |
225 |
const uint32_t cbp, |
const uint32_t cbp, |
226 |
Bitstream * bs, |
Bitstream * bs, |
227 |
const uint32_t quant, |
const uint32_t quant, |
228 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold, |
229 |
|
const unsigned int bound) |
230 |
{ |
{ |
231 |
|
|
232 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
245 |
|
|
246 |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
247 |
|
|
248 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
249 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
250 |
int16_t predictors[8]; |
int16_t predictors[8]; |
251 |
int start_coeff; |
int start_coeff; |
252 |
|
|
253 |
start_timer(); |
start_timer(); |
254 |
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], |
255 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound); |
256 |
{ |
if (!acpred_flag) { |
257 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
258 |
} |
} |
259 |
stop_prediction_timer(); |
stop_prediction_timer(); |
260 |
|
|
261 |
if (quant < intra_dc_threshold) |
if (quant < intra_dc_threshold) { |
|
{ |
|
262 |
int dc_size; |
int dc_size; |
263 |
int dc_dif; |
int dc_dif; |
264 |
|
|
265 |
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); |
266 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
267 |
|
|
268 |
if (dc_size > 8) |
if (dc_size > 8) { |
|
{ |
|
269 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); // marker |
270 |
} |
} |
271 |
|
|
272 |
block[i*64 + 0] = dc_dif; |
block[i*64 + 0] = dc_dif; |
273 |
start_coeff = 1; |
start_coeff = 1; |
274 |
} |
|
275 |
else |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
276 |
{ |
} else { |
277 |
start_coeff = 0; |
start_coeff = 0; |
278 |
} |
} |
279 |
|
|
280 |
start_timer(); |
start_timer(); |
281 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
282 |
{ |
{ |
283 |
get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
284 |
|
2 : pMB->acpred_directions[i]; |
285 |
|
|
286 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
287 |
} |
} |
288 |
stop_coding_timer(); |
stop_coding_timer(); |
289 |
|
|
292 |
stop_prediction_timer(); |
stop_prediction_timer(); |
293 |
|
|
294 |
start_timer(); |
start_timer(); |
295 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
296 |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
297 |
} |
} else { |
|
else |
|
|
{ |
|
298 |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
299 |
} |
} |
300 |
stop_iquant_timer(); |
stop_iquant_timer(); |
304 |
stop_idct_timer(); |
stop_idct_timer(); |
305 |
} |
} |
306 |
|
|
307 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
308 |
next_block = stride; |
next_block = stride; |
309 |
stride *= 2; |
stride *= 2; |
310 |
} |
} |
331 |
|
|
332 |
// decode an inter macroblock |
// decode an inter macroblock |
333 |
|
|
334 |
void decoder_mbinter(DECODER * dec, |
void |
335 |
|
decoder_mbinter(DECODER * dec, |
336 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
337 |
const uint32_t x_pos, |
const uint32_t x_pos, |
338 |
const uint32_t y_pos, |
const uint32_t y_pos, |
358 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
359 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
360 |
|
|
361 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
|
{ |
|
362 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
363 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
364 |
|
|
365 |
|
if (dec->quarterpel) |
366 |
|
{ |
367 |
|
uv_dx = (uv_dx >> 1) | (uv_dx & 1); |
368 |
|
uv_dy = (uv_dy >> 1) | (uv_dy & 1); |
369 |
|
} |
370 |
|
|
371 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
372 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
373 |
} |
} else { |
|
else |
|
|
{ |
|
374 |
int sum; |
int sum; |
375 |
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; |
376 |
|
|
377 |
|
if (dec->quarterpel) |
378 |
|
{ |
379 |
|
sum /= 2; |
380 |
|
} |
381 |
|
|
382 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
383 |
|
|
384 |
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; |
385 |
|
|
386 |
|
if (dec->quarterpel) |
387 |
|
{ |
388 |
|
sum /= 2; |
389 |
|
} |
390 |
|
|
391 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
392 |
} |
} |
393 |
|
|
394 |
start_timer(); |
start_timer(); |
395 |
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); |
if(dec->quarterpel) { |
396 |
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); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
397 |
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); |
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
398 |
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); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
399 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
400 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
401 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
402 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
403 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
404 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
405 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
406 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
407 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
408 |
|
} |
409 |
|
else { |
410 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
411 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
412 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
413 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
414 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
415 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
416 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
417 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
418 |
|
} |
419 |
|
|
420 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
421 |
|
uv_dx, uv_dy, stride2, rounding); |
422 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
423 |
|
uv_dx, uv_dy, stride2, rounding); |
424 |
stop_comp_timer(); |
stop_comp_timer(); |
425 |
|
|
426 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
427 |
{ |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
428 |
|
|
429 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
430 |
{ |
{ |
431 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
432 |
|
|
433 |
start_timer(); |
start_timer(); |
434 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, &block[i * 64], direction); |
435 |
stop_coding_timer(); |
stop_coding_timer(); |
436 |
|
|
437 |
start_timer(); |
start_timer(); |
438 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
439 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
440 |
} |
} else { |
|
else |
|
|
{ |
|
441 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
442 |
} |
} |
443 |
stop_iquant_timer(); |
stop_iquant_timer(); |
448 |
} |
} |
449 |
} |
} |
450 |
|
|
451 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
452 |
next_block = stride; |
next_block = stride; |
453 |
stride *= 2; |
stride *= 2; |
454 |
} |
} |
470 |
} |
} |
471 |
|
|
472 |
|
|
473 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
void |
474 |
|
decoder_iframe(DECODER * dec, |
475 |
|
Bitstream * bs, |
476 |
|
int quant, |
477 |
|
int intra_dc_threshold) |
478 |
{ |
{ |
479 |
|
uint32_t bound; |
480 |
uint32_t x, y; |
uint32_t x, y; |
481 |
|
|
482 |
for (y = 0; y < dec->mb_height; y++) |
bound = 0; |
|
{ |
|
|
for (x = 0; x < dec->mb_width; x++) |
|
|
{ |
|
|
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
|
483 |
|
|
484 |
|
for (y = 0; y < dec->mb_height; y++) { |
485 |
|
for (x = 0; x < dec->mb_width; x++) { |
486 |
|
MACROBLOCK *mb; |
487 |
uint32_t mcbpc; |
uint32_t mcbpc; |
488 |
uint32_t cbpc; |
uint32_t cbpc; |
489 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
490 |
uint32_t cbpy; |
uint32_t cbpy; |
491 |
uint32_t cbp; |
uint32_t cbp; |
492 |
|
|
493 |
|
while (BitstreamShowBits(bs, 9) == 1) |
494 |
|
BitstreamSkip(bs, 9); |
495 |
|
|
496 |
|
if (check_resync_marker(bs, 0)) |
497 |
|
{ |
498 |
|
bound = read_video_packet_header(bs, 0, &quant); |
499 |
|
x = bound % dec->mb_width; |
500 |
|
y = bound / dec->mb_width; |
501 |
|
} |
502 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
503 |
|
|
504 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
505 |
|
|
506 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
507 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
508 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
509 |
|
|
510 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
511 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
512 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
513 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
514 |
|
|
515 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
516 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
517 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
518 |
quant = 31; |
quant = 31; |
519 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
520 |
quant = 1; |
quant = 1; |
521 |
} |
} |
522 |
} |
} |
523 |
mb->quant = quant; |
mb->quant = quant; |
524 |
|
mb->mvs[0].x = mb->mvs[0].y = |
525 |
|
mb->mvs[1].x = mb->mvs[1].y = |
526 |
|
mb->mvs[2].x = mb->mvs[2].y = |
527 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
528 |
|
|
529 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
530 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
531 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DEBUG1("deci: field_dct: ", mb->field_dct); |
532 |
} |
} |
533 |
|
|
534 |
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, |
535 |
|
intra_dc_threshold, bound); |
536 |
} |
} |
537 |
|
if(dec->out_frm) |
538 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); |
539 |
|
|
540 |
} |
} |
541 |
|
|
542 |
} |
} |
543 |
|
|
544 |
|
|
545 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
void |
546 |
|
get_motion_vector(DECODER * dec, |
547 |
|
Bitstream * bs, |
548 |
|
int x, |
549 |
|
int y, |
550 |
|
int k, |
551 |
|
VECTOR * mv, |
552 |
|
int fcode, |
553 |
|
const int bound) |
554 |
{ |
{ |
555 |
|
|
556 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
558 |
int low = ((-32) * scale_fac); |
int low = ((-32) * scale_fac); |
559 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
560 |
|
|
561 |
VECTOR pmv[4]; |
VECTOR pmv; |
|
int32_t psad[4]; |
|
|
|
|
562 |
int mv_x, mv_y; |
int mv_x, mv_y; |
|
int pmv_x, pmv_y; |
|
563 |
|
|
564 |
|
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
|
get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); |
|
|
|
|
|
pmv_x = pmv[0].x; |
|
|
pmv_y = pmv[0].y; |
|
565 |
|
|
566 |
mv_x = get_mv(bs, fcode); |
mv_x = get_mv(bs, fcode); |
567 |
mv_y = get_mv(bs, fcode); |
mv_y = get_mv(bs, fcode); |
568 |
|
|
569 |
mv_x += pmv_x; |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); |
|
mv_y += pmv_y; |
|
570 |
|
|
571 |
if (mv_x < low) |
mv_x += pmv.x; |
572 |
{ |
mv_y += pmv.y; |
573 |
|
|
574 |
|
if (mv_x < low) { |
575 |
mv_x += range; |
mv_x += range; |
576 |
} |
} else if (mv_x > high) { |
|
else if (mv_x > high) |
|
|
{ |
|
577 |
mv_x -= range; |
mv_x -= range; |
578 |
} |
} |
579 |
|
|
580 |
if (mv_y < low) |
if (mv_y < low) { |
|
{ |
|
581 |
mv_y += range; |
mv_y += range; |
582 |
} |
} else if (mv_y > high) { |
|
else if (mv_y > high) |
|
|
{ |
|
583 |
mv_y -= range; |
mv_y -= range; |
584 |
} |
} |
585 |
|
|
589 |
} |
} |
590 |
|
|
591 |
|
|
592 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
void |
593 |
|
decoder_pframe(DECODER * dec, |
594 |
|
Bitstream * bs, |
595 |
|
int rounding, |
596 |
|
int quant, |
597 |
|
int fcode, |
598 |
|
int intra_dc_threshold) |
599 |
{ |
{ |
600 |
|
|
601 |
uint32_t x, y; |
uint32_t x, y; |
602 |
|
uint32_t bound; |
603 |
|
int cp_mb, st_mb; |
604 |
|
|
605 |
start_timer(); |
start_timer(); |
606 |
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, |
607 |
|
dec->width, dec->height); |
608 |
stop_edges_timer(); |
stop_edges_timer(); |
609 |
|
|
610 |
for (y = 0; y < dec->mb_height; y++) |
bound = 0; |
611 |
{ |
|
612 |
for (x = 0; x < dec->mb_width; x++) |
for (y = 0; y < dec->mb_height; y++) { |
613 |
|
cp_mb = st_mb = 0; |
614 |
|
for (x = 0; x < dec->mb_width; x++) { |
615 |
|
MACROBLOCK *mb; |
616 |
|
|
617 |
|
// skip stuffing |
618 |
|
while (BitstreamShowBits(bs, 10) == 1) |
619 |
|
BitstreamSkip(bs, 10); |
620 |
|
|
621 |
|
if (check_resync_marker(bs, fcode - 1)) |
622 |
{ |
{ |
623 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
624 |
|
x = bound % dec->mb_width; |
625 |
|
y = bound / dec->mb_width; |
626 |
|
} |
627 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
628 |
|
|
629 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
630 |
|
|
631 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
632 |
if (!(BitstreamGetBit(bs))) // not_coded |
if (!(BitstreamGetBit(bs))) // not_coded |
638 |
uint32_t cbp; |
uint32_t cbp; |
639 |
uint32_t intra; |
uint32_t intra; |
640 |
|
|
641 |
|
cp_mb++; |
642 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
643 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
644 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
645 |
|
|
646 |
|
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
647 |
|
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
648 |
acpred_flag = 0; |
acpred_flag = 0; |
649 |
|
|
650 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
651 |
|
|
652 |
if (intra) |
if (intra) { |
|
{ |
|
653 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
654 |
} |
} |
655 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
656 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
657 |
|
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
658 |
|
|
659 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
660 |
|
|
661 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
662 |
{ |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
663 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
664 |
if (quant > 31) |
quant += dquant; |
665 |
{ |
if (quant > 31) { |
666 |
quant = 31; |
quant = 31; |
667 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
668 |
quant = 1; |
quant = 1; |
669 |
} |
} |
670 |
|
DPRINTF(DPRINTF_MB, "quant %i", quant); |
671 |
} |
} |
672 |
mb->quant = quant; |
mb->quant = quant; |
673 |
|
|
674 |
if (dec->interlacing) |
if (dec->interlacing) { |
675 |
{ |
if (cbp || intra) { |
676 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
677 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DEBUG1("decp: field_dct: ", mb->field_dct); |
678 |
|
} |
679 |
|
|
680 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
681 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
682 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DEBUG1("decp: field_pred: ", mb->field_pred); |
683 |
|
|
684 |
if (mb->field_pred) |
if (mb->field_pred) { |
|
{ |
|
685 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
686 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
687 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
690 |
} |
} |
691 |
} |
} |
692 |
|
|
693 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
694 |
{ |
if (dec->interlacing && mb->field_pred) { |
695 |
if (dec->interlacing && mb->field_pred) |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
696 |
{ |
fcode, bound); |
697 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
698 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); |
fcode, bound); |
699 |
} |
} else { |
700 |
else |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
701 |
{ |
fcode, bound); |
702 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
703 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
mb->mvs[0].x; |
704 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
705 |
} |
mb->mvs[0].y; |
706 |
} |
} |
707 |
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
} else if (mb->mode == MODE_INTER4V ) { |
708 |
{ |
|
709 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
710 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
711 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
712 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
713 |
} |
} else // MODE_INTRA, MODE_INTRA_Q |
714 |
else // MODE_INTRA, MODE_INTRA_Q |
{ |
715 |
{ |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
716 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
0; |
717 |
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 = |
718 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
0; |
719 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
720 |
|
intra_dc_threshold, bound); |
721 |
continue; |
continue; |
722 |
} |
} |
723 |
|
|
724 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
725 |
} |
rounding); |
726 |
else // not coded |
} else // not coded |
727 |
{ |
{ |
728 |
//DEBUG2("P-frame MB at (X,Y)=",x,y); |
DEBUG2("P-frame MB at (X,Y)=",x,y); |
729 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
730 |
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; |
731 |
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; |
734 |
|
|
735 |
start_timer(); |
start_timer(); |
736 |
|
|
737 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
738 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
(16 * x), |
739 |
dec->edged_width); |
dec->refn[0].y + (16 * y) * dec->edged_width + |
740 |
|
(16 * x), dec->edged_width); |
741 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
|
742 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x+8), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
743 |
dec->edged_width); |
(16 * x + 8), |
744 |
|
dec->refn[0].y + (16 * y) * dec->edged_width + |
745 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
(16 * x + 8), dec->edged_width); |
746 |
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x), |
|
747 |
dec->edged_width); |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
748 |
|
(16 * x), |
749 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
dec->refn[0].y + (16 * y + |
750 |
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x+8), |
8) * dec->edged_width + |
751 |
dec->edged_width); |
(16 * x), dec->edged_width); |
752 |
|
|
753 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
754 |
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
(16 * x + 8), |
755 |
|
dec->refn[0].y + (16 * y + |
756 |
|
8) * dec->edged_width + |
757 |
|
(16 * x + 8), dec->edged_width); |
758 |
|
|
759 |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
760 |
|
(8 * x), |
761 |
|
dec->refn[0].u + |
762 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
763 |
dec->edged_width/2); |
dec->edged_width/2); |
764 |
|
|
765 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
766 |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
(8 * x), |
767 |
|
dec->refn[0].v + |
768 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
769 |
dec->edged_width/2); |
dec->edged_width/2); |
|
|
|
770 |
stop_transfer_timer(); |
stop_transfer_timer(); |
771 |
|
if(dec->out_frm && cp_mb > 0) { |
772 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
773 |
|
cp_mb = 0; |
774 |
} |
} |
775 |
|
st_mb = x+1; |
776 |
} |
} |
777 |
} |
} |
778 |
|
if(dec->out_frm && cp_mb > 0) |
779 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
780 |
|
} |
781 |
} |
} |
782 |
|
|
783 |
|
|
784 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
785 |
// decode B-frame motion vector |
// decode B-frame motion vector |
786 |
void get_b_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, VECTOR * mv, int fcode, const VECTOR pmv) |
void |
787 |
|
get_b_motion_vector(DECODER * dec, |
788 |
|
Bitstream * bs, |
789 |
|
int x, |
790 |
|
int y, |
791 |
|
VECTOR * mv, |
792 |
|
int fcode, |
793 |
|
const VECTOR pmv) |
794 |
{ |
{ |
795 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
796 |
int high = (32 * scale_fac) - 1; |
int high = (32 * scale_fac) - 1; |
809 |
mv_x += pmv_x; |
mv_x += pmv_x; |
810 |
mv_y += pmv_y; |
mv_y += pmv_y; |
811 |
|
|
812 |
if (mv_x < low) |
if (mv_x < low) { |
|
{ |
|
813 |
mv_x += range; |
mv_x += range; |
814 |
} |
} else if (mv_x > high) { |
|
else if (mv_x > high) |
|
|
{ |
|
815 |
mv_x -= range; |
mv_x -= range; |
816 |
} |
} |
817 |
|
|
818 |
if (mv_y < low) |
if (mv_y < low) { |
|
{ |
|
819 |
mv_y += range; |
mv_y += range; |
820 |
} |
} else if (mv_y > high) { |
|
else if (mv_y > high) |
|
|
{ |
|
821 |
mv_y -= range; |
mv_y -= range; |
822 |
} |
} |
823 |
|
|
828 |
|
|
829 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
830 |
// decode an B-frame forward & backward inter macroblock |
// decode an B-frame forward & backward inter macroblock |
831 |
void decoder_bf_mbinter(DECODER * dec, |
void |
832 |
|
decoder_bf_mbinter(DECODER * dec, |
833 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
834 |
const uint32_t x_pos, |
const uint32_t x_pos, |
835 |
const uint32_t y_pos, |
const uint32_t y_pos, |
854 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
855 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
856 |
|
|
857 |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
858 |
{ |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
859 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
860 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
861 |
|
|
862 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
863 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
864 |
} |
} else { |
|
else |
|
|
{ |
|
865 |
int sum; |
int sum; |
866 |
|
|
867 |
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; |
868 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = |
869 |
|
(sum == |
870 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
871 |
|
(ABS(sum) / 16) * 2)); |
872 |
|
|
873 |
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; |
874 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = |
875 |
|
(sum == |
876 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
877 |
|
(ABS(sum) / 16) * 2)); |
878 |
} |
} |
879 |
|
|
880 |
start_timer(); |
start_timer(); |
881 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos, |
882 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
883 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
884 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
885 |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
886 |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
887 |
|
0); |
888 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
889 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
890 |
|
0); |
891 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
892 |
|
uv_dx, uv_dy, stride2, 0); |
893 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
894 |
|
uv_dx, uv_dy, stride2, 0); |
895 |
stop_comp_timer(); |
stop_comp_timer(); |
896 |
|
|
897 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
898 |
{ |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
899 |
|
|
900 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
901 |
{ |
{ |
902 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
903 |
|
|
904 |
start_timer(); |
start_timer(); |
905 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, &block[i * 64], direction); |
906 |
stop_coding_timer(); |
stop_coding_timer(); |
907 |
|
|
908 |
start_timer(); |
start_timer(); |
909 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
910 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
911 |
} |
} else { |
|
else |
|
|
{ |
|
912 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
913 |
} |
} |
914 |
stop_iquant_timer(); |
stop_iquant_timer(); |
919 |
} |
} |
920 |
} |
} |
921 |
|
|
922 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
923 |
next_block = stride; |
next_block = stride; |
924 |
stride *= 2; |
stride *= 2; |
925 |
} |
} |
943 |
|
|
944 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
945 |
// decode an B-frame direct & inter macroblock |
// decode an B-frame direct & inter macroblock |
946 |
void decoder_bf_interpolate_mbinter(DECODER * dec, |
void |
947 |
IMAGE forward, IMAGE backward, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
948 |
|
IMAGE forward, |
949 |
|
IMAGE backward, |
950 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
951 |
const uint32_t x_pos, |
const uint32_t x_pos, |
952 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t cbp, |
|
953 |
Bitstream * bs ) |
Bitstream * bs ) |
954 |
{ |
{ |
955 |
|
|
964 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
965 |
uint32_t i; |
uint32_t i; |
966 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
967 |
|
const uint32_t cbp = pMB->cbp; |
968 |
|
|
969 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
970 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
971 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
972 |
|
|
973 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
974 |
{ |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
975 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
976 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
977 |
|
|
983 |
|
|
984 |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
985 |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
986 |
} |
} else { |
|
else |
|
|
{ |
|
987 |
int sum; |
int sum; |
988 |
|
|
989 |
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; |
990 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = |
991 |
|
(sum == |
992 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
993 |
|
(ABS(sum) / 16) * 2)); |
994 |
|
|
995 |
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; |
996 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = |
997 |
|
(sum == |
998 |
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
999 |
b_uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
(ABS(sum) / 16) * 2)); |
1000 |
|
|
1001 |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
sum = |
1002 |
b_uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
1003 |
} |
pMB->b_mvs[3].x; |
1004 |
|
b_uv_dx = |
1005 |
|
(sum == |
1006 |
start_timer(); |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1007 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
(ABS(sum) / 16) * 2)); |
1008 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
|
1009 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
sum = |
1010 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
1011 |
interpolate8x8_switch(dec->cur.u, forward.u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
pMB->b_mvs[3].y; |
1012 |
interpolate8x8_switch(dec->cur.v, forward.v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
b_uv_dy = |
1013 |
|
(sum == |
1014 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1015 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos, 16*y_pos , pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
(ABS(sum) / 16) * 2)); |
1016 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos + 8, 16*y_pos , pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
} |
1017 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos, 16*y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
|
1018 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos + 8, 16*y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
|
1019 |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8*x_pos, 8*y_pos, b_uv_dx, b_uv_dy, stride2, 0); |
start_timer(); |
1020 |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8*x_pos, 8*y_pos, b_uv_dx, b_uv_dy, stride2, 0); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1021 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1022 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos, 16*y_pos , stride); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1023 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos + 8, 16*y_pos , stride); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1024 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos, 16*y_pos + 8, stride); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1025 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos + 8, 16*y_pos + 8, stride); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1026 |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8*x_pos, 8*y_pos, stride2); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1027 |
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8*x_pos, 8*y_pos, stride2); |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1028 |
|
0); |
1029 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1030 |
|
uv_dy, stride2, 0); |
1031 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1032 |
|
uv_dy, stride2, 0); |
1033 |
|
|
1034 |
|
|
1035 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
1036 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1037 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1038 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1039 |
|
0); |
1040 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, |
1041 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1042 |
|
stride, 0); |
1043 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1044 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1045 |
|
stride, 0); |
1046 |
|
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
1047 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1048 |
|
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1049 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1050 |
|
|
1051 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1052 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1053 |
|
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos, |
1054 |
|
stride, 0); |
1055 |
|
|
1056 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1057 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1058 |
|
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1059 |
|
stride, 0); |
1060 |
|
|
1061 |
|
interpolate8x8_avg2(dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos, |
1062 |
|
dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos, |
1063 |
|
dec->refn[2].y + (16 * (y_pos + 8) * stride) + 16 * x_pos, |
1064 |
|
stride, 0); |
1065 |
|
|
1066 |
|
interpolate8x8_avg2(dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos + 8, |
1067 |
|
dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos + 8, |
1068 |
|
dec->refn[2].y + (16 * (y_pos + 8) * stride) + 16 * x_pos + 8, |
1069 |
|
stride, 0); |
1070 |
|
|
1071 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride) + 8 * x_pos, |
1072 |
|
dec->cur.u + (8 * y_pos * stride) + 8 * x_pos, |
1073 |
|
dec->refn[2].u + (8 * y_pos * stride) + 8 * x_pos, |
1074 |
|
stride2, 0); |
1075 |
|
|
1076 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride) + 8 * x_pos, |
1077 |
|
dec->cur.v + (8 * y_pos * stride) + 8 * x_pos, |
1078 |
|
dec->refn[2].v + (8 * y_pos * stride) + 8 * x_pos, |
1079 |
|
stride2, 0); |
1080 |
|
|
1081 |
stop_comp_timer(); |
stop_comp_timer(); |
1082 |
|
|
1083 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
1084 |
{ |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1085 |
|
|
1086 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
1087 |
{ |
{ |
1088 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
1089 |
|
|
1090 |
start_timer(); |
start_timer(); |
1091 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, &block[i * 64], direction); |
1092 |
stop_coding_timer(); |
stop_coding_timer(); |
1093 |
|
|
1094 |
start_timer(); |
start_timer(); |
1095 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
1096 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
1097 |
} |
} else { |
|
else |
|
|
{ |
|
1098 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
1099 |
} |
} |
1100 |
stop_iquant_timer(); |
stop_iquant_timer(); |
1105 |
} |
} |
1106 |
} |
} |
1107 |
|
|
1108 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
1109 |
next_block = stride; |
next_block = stride; |
1110 |
stride *= 2; |
stride *= 2; |
1111 |
} |
} |
1129 |
|
|
1130 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
1131 |
// for decode B-frame dbquant |
// for decode B-frame dbquant |
1132 |
int32_t __inline get_dbquant(Bitstream * bs) |
int32_t __inline |
1133 |
|
get_dbquant(Bitstream * bs) |
1134 |
{ |
{ |
1135 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) // '0' |
1136 |
return(0); |
return(0); |
1147 |
// 01 1 |
// 01 1 |
1148 |
// 001 2 |
// 001 2 |
1149 |
// 0001 3 |
// 0001 3 |
1150 |
int32_t __inline get_mbtype(Bitstream * bs) |
int32_t __inline |
1151 |
|
get_mbtype(Bitstream * bs) |
1152 |
{ |
{ |
1153 |
int32_t mb_type; |
int32_t mb_type; |
1154 |
|
|
1155 |
for(mb_type=0;mb_type<=3;mb_type++){ |
for(mb_type=0;mb_type<=3;mb_type++){ |
1156 |
if (BitstreamGetBit(bs)) |
if (BitstreamGetBit(bs)) |
1157 |
break; |
break; |
1163 |
return(-1); |
return(-1); |
1164 |
} |
} |
1165 |
|
|
1166 |
void decoder_bframe(DECODER * dec, Bitstream * bs, int quant, int fcode_forward, int fcode_backward) |
void |
1167 |
|
decoder_bframe(DECODER * dec, |
1168 |
|
Bitstream * bs, |
1169 |
|
int quant, |
1170 |
|
int fcode_forward, |
1171 |
|
int fcode_backward) |
1172 |
{ |
{ |
|
|
|
1173 |
uint32_t x, y; |
uint32_t x, y; |
1174 |
VECTOR mv, zeromv; |
VECTOR mv; |
1175 |
|
const VECTOR zeromv = {0,0}; |
1176 |
|
#ifdef BFRAMES_DEC_DEBUG |
1177 |
|
FILE *fp; |
1178 |
|
static char first=0; |
1179 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1180 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1181 |
|
} |
1182 |
|
#endif |
1183 |
|
|
1184 |
start_timer(); |
start_timer(); |
1185 |
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, |
1186 |
//image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
dec->width, dec->height); |
1187 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1188 |
|
dec->width, dec->height); |
1189 |
stop_edges_timer(); |
stop_edges_timer(); |
1190 |
|
|
1191 |
|
#ifdef BFRAMES_DEC_DEBUG |
1192 |
|
if (!first){ |
1193 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1194 |
|
} |
1195 |
|
#endif |
1196 |
|
|
1197 |
for (y = 0; y < dec->mb_height; y++) |
for (y = 0; y < dec->mb_height; y++) { |
|
{ |
|
1198 |
// Initialize Pred Motion Vector |
// Initialize Pred Motion Vector |
1199 |
dec->p_fmv.x = dec->p_fmv.y = dec->p_bmv.x = dec->p_bmv.y = 0; |
dec->p_fmv = dec->p_bmv = zeromv; |
1200 |
for (x = 0; x < dec->mb_width; x++) |
for (x = 0; x < dec->mb_width; x++) { |
|
{ |
|
1201 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
1202 |
MACROBLOCK * last_mb = &dec->last_mbs[y*dec->mb_width + x]; |
MACROBLOCK * last_mb = &dec->last_mbs[y*dec->mb_width + x]; |
1203 |
|
|
1204 |
mb->mvs[0].x=mb->mvs[0].y=zeromv.x = zeromv.y = mv.x = mv.y = 0; |
mv = |
1205 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1206 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1207 |
|
|
1208 |
// the last P_VOP is skip macroblock ? |
// the last P_VOP is skip macroblock ? |
1209 |
if (last_mb->mode == MODE_NOT_CODED){ |
if (last_mb->mode == MODE_NOT_CODED){ |
1210 |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
|
mb->mb_type = MODE_FORWARD; |
|
1211 |
mb->cbp = 0; |
mb->cbp = 0; |
1212 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
#ifdef BFRAMES_DEC_DEBUG |
1213 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
mb->mb_type = MODE_NOT_CODED; |
1214 |
mb->quant = 8; |
BFRAME_DEBUG |
1215 |
|
#endif |
1216 |
|
mb->mb_type = MODE_FORWARD; |
1217 |
|
mb->quant = last_mb->quant; |
1218 |
|
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1219 |
|
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1220 |
|
|
1221 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1222 |
continue; |
continue; |
1223 |
} |
} |
1224 |
|
|
|
//t=BitstreamShowBits(bs,32); |
|
|
|
|
1225 |
if (!BitstreamGetBit(bs)){ // modb=='0' |
if (!BitstreamGetBit(bs)){ // modb=='0' |
1226 |
const uint8_t modb2=BitstreamGetBit(bs); |
const uint8_t modb2=BitstreamGetBit(bs); |
1227 |
|
|
1235 |
if (mb->mb_type && mb->cbp){ |
if (mb->mb_type && mb->cbp){ |
1236 |
quant += get_dbquant(bs); |
quant += get_dbquant(bs); |
1237 |
|
|
1238 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
1239 |
quant = 31; |
quant = 31; |
1240 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
1241 |
quant = 1; |
quant = 1; |
1242 |
} |
} |
|
} else { |
|
|
quant = 8; |
|
1243 |
} |
} |
|
mb->quant = quant; |
|
1244 |
} else { |
} else { |
1245 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mb_type = MODE_DIRECT_NONE_MV; |
1246 |
mb->cbp = 0; |
mb->cbp = 0; |
1247 |
} |
} |
1248 |
|
|
1249 |
mb->mode = MODE_INTER; |
mb->quant = quant; |
1250 |
|
mb->mode = MODE_INTER4V; |
1251 |
//DEBUG1("Switch bm_type=",mb->mb_type); |
//DEBUG1("Switch bm_type=",mb->mb_type); |
1252 |
|
|
1253 |
switch(mb->mb_type) |
#ifdef BFRAMES_DEC_DEBUG |
1254 |
{ |
BFRAME_DEBUG |
1255 |
|
#endif |
1256 |
|
|
1257 |
|
switch (mb->mb_type) { |
1258 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1259 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], 1, zeromv); |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1260 |
|
|
1261 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
1262 |
{ // Because this file is a C file not C++ so I use '{' to define var |
{ |
1263 |
const int64_t TRB=dec->time_pp-dec->time_bp, |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
TRD=dec->time_pp; |
|
1264 |
int i; |
int i; |
1265 |
|
|
1266 |
for(i=0;i<4;i++){ |
for(i=0;i<4;i++){ |
1267 |
mb->mvs[i].x = (int32_t)((TRB * last_mb->mvs[i].x)/TRD+mb->mvs[0].x); |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1268 |
mb->b_mvs[i].x = (int32_t)((mb->mvs[0].x==0)?((TRB-TRD)*last_mb->mvs[i].x)/TRD:mb->mvs[i].x-last_mb->mvs[i].x); |
/ TRD + mv.x); |
1269 |
mb->mvs[i].y = (int32_t)((TRB * last_mb->mvs[i].y)/TRD+mb->mvs[0].y); |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1270 |
mb->b_mvs[i].y = (int32_t)((mb->mvs[0].y==0)?((TRB-TRD)*last_mb->mvs[i].y)/TRD:mb->mvs[i].y-last_mb->mvs[i].y); |
? ((TRB - TRD) * last_mb->mvs[i].x) |
1271 |
|
/ TRD |
1272 |
|
: mb->mvs[i].x - last_mb->mvs[i].x); |
1273 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1274 |
|
/ TRD + mv.y); |
1275 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1276 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
1277 |
|
/ TRD |
1278 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1279 |
} |
} |
1280 |
//DEBUG("B-frame Direct!\n"); |
//DEBUG("B-frame Direct!\n"); |
1281 |
} |
} |
1282 |
mb->mode = MODE_INTER4V; |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1283 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, mb->cbp, bs); |
mb, x, y, bs); |
1284 |
break; |
break; |
1285 |
|
|
1286 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1287 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1288 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_fmv); |
1289 |
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1290 |
|
|
1291 |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], fcode_backward, dec->p_bmv); |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1292 |
dec->p_bmv.x = mb->b_mvs[1].x = mb->b_mvs[2].x = mb->b_mvs[3].x = mb->b_mvs[0].x; |
fcode_backward, dec->p_bmv); |
1293 |
dec->p_bmv.y = mb->b_mvs[1].y = mb->b_mvs[2].y = mb->b_mvs[3].y = mb->b_mvs[0].y; |
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1294 |
|
mb->b_mvs[3] = mb->b_mvs[0]; |
1295 |
|
|
1296 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, mb->cbp, bs); |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1297 |
|
mb, x, y, bs); |
1298 |
//DEBUG("B-frame Bidir!\n"); |
//DEBUG("B-frame Bidir!\n"); |
1299 |
break; |
break; |
1300 |
|
|
1301 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1302 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, dec->p_bmv); |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1303 |
dec->p_bmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_bmv); |
1304 |
dec->p_bmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1305 |
|
|
1306 |
|
mb->mode = MODE_INTER; |
1307 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1308 |
//DEBUG("B-frame Backward!\n"); |
//DEBUG("B-frame Backward!\n"); |
1309 |
break; |
break; |
1310 |
|
|
1311 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1312 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1313 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_fmv); |
1314 |
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1315 |
|
|
1316 |
|
mb->mode = MODE_INTER; |
1317 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1318 |
//DEBUG("B-frame Forward!\n"); |
//DEBUG("B-frame Forward!\n"); |
1319 |
break; |
break; |
1324 |
|
|
1325 |
} // end of FOR |
} // end of FOR |
1326 |
} |
} |
1327 |
|
#ifdef BFRAMES_DEC_DEBUG |
1328 |
|
if (!first){ |
1329 |
|
first=1; |
1330 |
|
if (fp) |
1331 |
|
fclose(fp); |
1332 |
|
} |
1333 |
|
#endif |
1334 |
} |
} |
1335 |
|
|
1336 |
// swap two MACROBLOCK array |
// swap two MACROBLOCK array |
1337 |
void mb_swap(MACROBLOCK **mb1, MACROBLOCK **mb2) |
void |
1338 |
|
mb_swap(MACROBLOCK ** mb1, |
1339 |
|
MACROBLOCK ** mb2) |
1340 |
{ |
{ |
1341 |
MACROBLOCK *temp=*mb1; |
MACROBLOCK *temp=*mb1; |
1342 |
|
|
1343 |
*mb1=*mb2; |
*mb1=*mb2; |
1344 |
*mb2=temp; |
*mb2=temp; |
1345 |
} |
} |
1346 |
|
|
1347 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
int |
1348 |
|
decoder_decode(DECODER * dec, |
1349 |
|
XVID_DEC_FRAME * frame) |
1350 |
{ |
{ |
1351 |
|
|
1352 |
Bitstream bs; |
Bitstream bs; |
1359 |
|
|
1360 |
start_global_timer(); |
start_global_timer(); |
1361 |
|
|
1362 |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1363 |
|
|
1364 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1365 |
|
|
1366 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
1367 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
1368 |
dec->frames++; |
dec->frames++; |
1369 |
vop_type=BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold); |
vop_type = |
1370 |
|
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
1371 |
|
&fcode_backward, &intra_dc_threshold); |
1372 |
|
|
1373 |
dec->p_bmv.x=dec->p_bmv.y=dec->p_fmv.y=dec->p_fmv.y=0; // init pred vector to 0 |
dec->p_bmv.x=dec->p_bmv.y=dec->p_fmv.y=dec->p_fmv.y=0; // init pred vector to 0 |
1374 |
|
|
1375 |
switch (vop_type) |
switch (vop_type) { |
|
{ |
|
1376 |
case P_VOP : |
case P_VOP : |
1377 |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, intra_dc_threshold); |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1378 |
|
intra_dc_threshold); |
1379 |
|
#ifdef BFRAMES_DEC |
1380 |
DEBUG1("P_VOP Time=",dec->time); |
DEBUG1("P_VOP Time=",dec->time); |
1381 |
|
#endif |
1382 |
break; |
break; |
1383 |
|
|
1384 |
case I_VOP : |
case I_VOP : |
1385 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1386 |
|
#ifdef BFRAMES_DEC |
1387 |
DEBUG1("I_VOP Time=",dec->time); |
DEBUG1("I_VOP Time=",dec->time); |
1388 |
|
#endif |
1389 |
break; |
break; |
1390 |
|
|
1391 |
case B_VOP : |
case B_VOP : |
1392 |
#ifdef BFRAMES |
#ifdef BFRAMES_DEC |
1393 |
if (dec->time_pp > dec->time_bp){ |
if (dec->time_pp > dec->time_bp){ |
1394 |
DEBUG1("B_VOP Time=",dec->time); |
DEBUG1("B_VOP Time=",dec->time); |
1395 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1396 |
} else { |
} else { |
1397 |
DEBUG("broken B-frame!"); |
DEBUG("broken B-frame!"); |
1398 |
} |
} |
1399 |
|
#else |
1400 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1401 |
#endif |
#endif |
1402 |
break; |
break; |
1403 |
|
|
1404 |
case N_VOP : // vop not coded |
case N_VOP : // vop not coded |
1405 |
|
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
1406 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1407 |
break; |
break; |
1408 |
|
|
1409 |
default : |
default : |
1410 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1411 |
} |
} |
1412 |
|
|
1413 |
|
#ifdef BFRAMES_DEC_DEBUG |
1414 |
|
if (frame->length != BitstreamPos(&bs) / 8){ |
1415 |
|
DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8); |
1416 |
|
} |
1417 |
|
#endif |
1418 |
frame->length = BitstreamPos(&bs) / 8; |
frame->length = BitstreamPos(&bs) / 8; |
1419 |
|
|
1420 |
#ifdef BFRAMES |
|
1421 |
|
#ifdef BFRAMES_DEC |
1422 |
// test if no B_VOP |
// test if no B_VOP |
1423 |
if (dec->low_delay){ |
if (dec->low_delay || dec->frames == 0) { |
1424 |
#endif |
#endif |
1425 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1426 |
frame->image, frame->stride, frame->colorspace); |
frame->image, frame->stride, frame->colorspace); |
1427 |
#ifdef BFRAMES |
|
1428 |
|
#ifdef BFRAMES_DEC |
1429 |
} else { |
} else { |
1430 |
if (dec->frames >= 1){ |
if (dec->frames >= 1){ |
1431 |
start_timer(); |
start_timer(); |
1432 |
if ((vop_type == I_VOP || vop_type == P_VOP)) |
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
1433 |
{ |
image_output(&dec->refn[0], dec->width, dec->height, |
1434 |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
dec->edged_width, frame->image, frame->stride, |
1435 |
frame->image, frame->stride, frame->colorspace); |
frame->colorspace); |
1436 |
} else if (vop_type == B_VOP) { |
} else if (vop_type == B_VOP) { |
1437 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
image_output(&dec->cur, dec->width, dec->height, |
1438 |
frame->image, frame->stride, frame->colorspace); |
dec->edged_width, frame->image, frame->stride, |
1439 |
|
frame->colorspace); |
1440 |
} |
} |
1441 |
stop_conv_timer(); |
stop_conv_timer(); |
1442 |
} |
} |
1446 |
if (vop_type==I_VOP || vop_type==P_VOP){ |
if (vop_type==I_VOP || vop_type==P_VOP){ |
1447 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1448 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1449 |
|
|
1450 |
// swap MACROBLOCK |
// swap MACROBLOCK |
1451 |
if (dec->low_delay && vop_type==P_VOP) |
// the Divx will not set the low_delay flage some times |
1452 |
|
// so follow code will wrong to not swap at that time |
1453 |
|
// this will broken bitstream! so I'm change it, |
1454 |
|
// But that is not the best way! can anyone tell me how |
1455 |
|
// to do another way? |
1456 |
|
// 18-07-2002 MinChen<chenm001@163.com> |
1457 |
|
//if (!dec->low_delay && vop_type == P_VOP) |
1458 |
|
if (vop_type == P_VOP) |
1459 |
mb_swap(&dec->mbs, &dec->last_mbs); |
mb_swap(&dec->mbs, &dec->last_mbs); |
1460 |
} |
} |
1461 |
|
|