39 |
#include "../bitstream/zigzag.h" |
#include "../bitstream/zigzag.h" |
40 |
#include "../dct/fdct.h" |
#include "../dct/fdct.h" |
41 |
#include "../dct/idct.h" |
#include "../dct/idct.h" |
42 |
#include "../quant/quant_mpeg4.h" |
#include "../quant/quant.h" |
|
#include "../quant/quant_h263.h" |
|
43 |
#include "../encoder.h" |
#include "../encoder.h" |
44 |
|
|
45 |
#include "../image/reduced.h" |
#include "../image/reduced.h" |
46 |
|
#include "../quant/quant_matrix.h" |
47 |
|
|
48 |
MBFIELDTEST_PTR MBFieldTest; |
MBFIELDTEST_PTR MBFieldTest; |
49 |
|
|
126 |
int mpeg; |
int mpeg; |
127 |
int scaler_lum, scaler_chr; |
int scaler_lum, scaler_chr; |
128 |
|
|
129 |
quanth263_intraFuncPtr const quant[2] = |
quant_intraFuncPtr const quant[2] = |
130 |
{ |
{ |
131 |
(quanth263_intraFuncPtr)quant_intra, |
quant_h263_intra, |
132 |
(quanth263_intraFuncPtr)quant4_intra |
quant_mpeg_intra |
133 |
}; |
}; |
134 |
|
|
135 |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
138 |
|
|
139 |
/* Quantize the block */ |
/* Quantize the block */ |
140 |
start_timer(); |
start_timer(); |
141 |
quant[mpeg](&data[0 * 64], &qcoeff[0 * 64], pMB->quant, scaler_lum); |
quant[mpeg](&data[0 * 64], &qcoeff[0 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices); |
142 |
quant[mpeg](&data[1 * 64], &qcoeff[1 * 64], pMB->quant, scaler_lum); |
quant[mpeg](&data[1 * 64], &qcoeff[1 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices); |
143 |
quant[mpeg](&data[2 * 64], &qcoeff[2 * 64], pMB->quant, scaler_lum); |
quant[mpeg](&data[2 * 64], &qcoeff[2 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices); |
144 |
quant[mpeg](&data[3 * 64], &qcoeff[3 * 64], pMB->quant, scaler_lum); |
quant[mpeg](&data[3 * 64], &qcoeff[3 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices); |
145 |
quant[mpeg](&data[4 * 64], &qcoeff[4 * 64], pMB->quant, scaler_chr); |
quant[mpeg](&data[4 * 64], &qcoeff[4 * 64], pMB->quant, scaler_chr, pParam->mpeg_quant_matrices); |
146 |
quant[mpeg](&data[5 * 64], &qcoeff[5 * 64], pMB->quant, scaler_chr); |
quant[mpeg](&data[5 * 64], &qcoeff[5 * 64], pMB->quant, scaler_chr, pParam->mpeg_quant_matrices); |
147 |
stop_quant_timer(); |
stop_quant_timer(); |
148 |
} |
} |
149 |
|
|
157 |
int mpeg; |
int mpeg; |
158 |
int scaler_lum, scaler_chr; |
int scaler_lum, scaler_chr; |
159 |
|
|
160 |
quanth263_intraFuncPtr const dequant[2] = |
quant_intraFuncPtr const dequant[2] = |
161 |
{ |
{ |
162 |
(quanth263_intraFuncPtr)dequant_intra, |
dequant_h263_intra, |
163 |
(quanth263_intraFuncPtr)dequant4_intra |
dequant_mpeg_intra |
164 |
}; |
}; |
165 |
|
|
166 |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
168 |
scaler_chr = get_dc_scaler(iQuant, 0); |
scaler_chr = get_dc_scaler(iQuant, 0); |
169 |
|
|
170 |
start_timer(); |
start_timer(); |
171 |
dequant[mpeg](&qcoeff[0 * 64], &data[0 * 64], iQuant, scaler_lum); |
dequant[mpeg](&qcoeff[0 * 64], &data[0 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices); |
172 |
dequant[mpeg](&qcoeff[1 * 64], &data[1 * 64], iQuant, scaler_lum); |
dequant[mpeg](&qcoeff[1 * 64], &data[1 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices); |
173 |
dequant[mpeg](&qcoeff[2 * 64], &data[2 * 64], iQuant, scaler_lum); |
dequant[mpeg](&qcoeff[2 * 64], &data[2 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices); |
174 |
dequant[mpeg](&qcoeff[3 * 64], &data[3 * 64], iQuant, scaler_lum); |
dequant[mpeg](&qcoeff[3 * 64], &data[3 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices); |
175 |
dequant[mpeg](&qcoeff[4 * 64], &data[4 * 64], iQuant, scaler_chr); |
dequant[mpeg](&qcoeff[4 * 64], &data[4 * 64], iQuant, scaler_chr, pParam->mpeg_quant_matrices); |
176 |
dequant[mpeg](&qcoeff[5 * 64], &data[5 * 64], iQuant, scaler_chr); |
dequant[mpeg](&qcoeff[5 * 64], &data[5 * 64], iQuant, scaler_chr, pParam->mpeg_quant_matrices); |
177 |
stop_iquant_timer(); |
stop_iquant_timer(); |
178 |
} |
} |
179 |
|
|
|
|
|
|
typedef int (*trellis_func_ptr_t)(int16_t *const Out, |
|
|
const int16_t *const In, |
|
|
int Q, |
|
|
const uint16_t * const Zigzag, |
|
|
int Non_Zero); |
|
|
|
|
|
static int |
|
|
dct_quantize_trellis_h263_c(int16_t *const Out, |
|
|
const int16_t *const In, |
|
|
int Q, |
|
|
const uint16_t * const Zigzag, |
|
|
int Non_Zero); |
|
|
|
|
180 |
static int |
static int |
181 |
dct_quantize_trellis_mpeg_c(int16_t *const Out, |
dct_quantize_trellis_c(int16_t *const Out, |
182 |
const int16_t *const In, |
const int16_t *const In, |
183 |
int Q, |
int Q, |
184 |
const uint16_t * const Zigzag, |
const uint16_t * const Zigzag, |
185 |
|
const uint16_t * const QuantMatrix, |
186 |
int Non_Zero); |
int Non_Zero); |
187 |
|
|
188 |
/* Quantize all blocks -- Inter mode */ |
/* Quantize all blocks -- Inter mode */ |
201 |
int sum; |
int sum; |
202 |
int code_block, mpeg; |
int code_block, mpeg; |
203 |
|
|
204 |
quanth263_interFuncPtr const quant[2] = |
quant_interFuncPtr const quant[2] = |
205 |
{ |
{ |
206 |
(quanth263_interFuncPtr)quant_inter, |
quant_h263_inter, |
207 |
(quanth263_interFuncPtr)quant4_inter |
quant_mpeg_inter |
|
}; |
|
|
|
|
|
trellis_func_ptr_t const trellis[2] = |
|
|
{ |
|
|
(trellis_func_ptr_t)dct_quantize_trellis_h263_c, |
|
|
(trellis_func_ptr_t)dct_quantize_trellis_mpeg_c |
|
208 |
}; |
}; |
209 |
|
|
210 |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
214 |
/* Quantize the block */ |
/* Quantize the block */ |
215 |
start_timer(); |
start_timer(); |
216 |
|
|
217 |
sum = quant[mpeg](&qcoeff[i*64], &data[i*64], pMB->quant); |
sum = quant[mpeg](&qcoeff[i*64], &data[i*64], pMB->quant, pParam->mpeg_quant_matrices); |
218 |
|
|
219 |
if(sum && (frame->vop_flags & XVID_VOP_TRELLISQUANT)) { |
if(sum && (frame->vop_flags & XVID_VOP_TRELLISQUANT)) { |
220 |
sum = trellis[mpeg](&qcoeff[i*64], &data[i*64], pMB->quant, &scan_tables[0][0], 63); |
const uint16_t *matrix; |
221 |
|
const static uint16_t h263matrix[] = |
222 |
|
{ |
223 |
|
16, 16, 16, 16, 16, 16, 16, 16, |
224 |
|
16, 16, 16, 16, 16, 16, 16, 16, |
225 |
|
16, 16, 16, 16, 16, 16, 16, 16, |
226 |
|
16, 16, 16, 16, 16, 16, 16, 16, |
227 |
|
16, 16, 16, 16, 16, 16, 16, 16, |
228 |
|
16, 16, 16, 16, 16, 16, 16, 16, |
229 |
|
16, 16, 16, 16, 16, 16, 16, 16, |
230 |
|
16, 16, 16, 16, 16, 16, 16, 16 |
231 |
|
}; |
232 |
|
|
233 |
|
matrix = (mpeg)?get_inter_matrix(pParam->mpeg_quant_matrices):h263matrix; |
234 |
|
sum = dct_quantize_trellis_c(&qcoeff[i*64], &data[i*64], |
235 |
|
pMB->quant, &scan_tables[0][0], |
236 |
|
matrix, |
237 |
|
63); |
238 |
} |
} |
239 |
stop_quant_timer(); |
stop_quant_timer(); |
240 |
|
|
275 |
{ |
{ |
276 |
int mpeg; |
int mpeg; |
277 |
|
|
278 |
quanth263_interFuncPtr const dequant[2] = |
quant_interFuncPtr const dequant[2] = |
279 |
{ |
{ |
280 |
(quanth263_interFuncPtr)dequant_inter, |
dequant_h263_inter, |
281 |
(quanth263_interFuncPtr)dequant4_inter |
dequant_mpeg_inter |
282 |
}; |
}; |
283 |
|
|
284 |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT); |
285 |
|
|
286 |
start_timer(); |
start_timer(); |
287 |
if(cbp & (1 << (5 - 0))) dequant[mpeg](&data[0 * 64], &qcoeff[0 * 64], iQuant); |
if(cbp & (1 << (5 - 0))) dequant[mpeg](&data[0 * 64], &qcoeff[0 * 64], iQuant, pParam->mpeg_quant_matrices); |
288 |
if(cbp & (1 << (5 - 1))) dequant[mpeg](&data[1 * 64], &qcoeff[1 * 64], iQuant); |
if(cbp & (1 << (5 - 1))) dequant[mpeg](&data[1 * 64], &qcoeff[1 * 64], iQuant, pParam->mpeg_quant_matrices); |
289 |
if(cbp & (1 << (5 - 2))) dequant[mpeg](&data[2 * 64], &qcoeff[2 * 64], iQuant); |
if(cbp & (1 << (5 - 2))) dequant[mpeg](&data[2 * 64], &qcoeff[2 * 64], iQuant, pParam->mpeg_quant_matrices); |
290 |
if(cbp & (1 << (5 - 3))) dequant[mpeg](&data[3 * 64], &qcoeff[3 * 64], iQuant); |
if(cbp & (1 << (5 - 3))) dequant[mpeg](&data[3 * 64], &qcoeff[3 * 64], iQuant, pParam->mpeg_quant_matrices); |
291 |
if(cbp & (1 << (5 - 4))) dequant[mpeg](&data[4 * 64], &qcoeff[4 * 64], iQuant); |
if(cbp & (1 << (5 - 4))) dequant[mpeg](&data[4 * 64], &qcoeff[4 * 64], iQuant, pParam->mpeg_quant_matrices); |
292 |
if(cbp & (1 << (5 - 5))) dequant[mpeg](&data[5 * 64], &qcoeff[5 * 64], iQuant); |
if(cbp & (1 << (5 - 5))) dequant[mpeg](&data[5 * 64], &qcoeff[5 * 64], iQuant, pParam->mpeg_quant_matrices); |
293 |
stop_iquant_timer(); |
stop_iquant_timer(); |
294 |
} |
} |
295 |
|
|
360 |
uint32_t cst; |
uint32_t cst; |
361 |
int vop_reduced; |
int vop_reduced; |
362 |
const IMAGE * const pCurrent = &frame->image; |
const IMAGE * const pCurrent = &frame->image; |
363 |
|
|
364 |
/* Array of function pointers, indexed by [vop_reduced<<1+add] */ |
/* Array of function pointers, indexed by [vop_reduced<<1+add] */ |
365 |
transfer_operation_16to8_t * const functions[4] = |
transfer_operation_16to8_t * const functions[4] = |
366 |
{ |
{ |
448 |
uint8_t cbp; |
uint8_t cbp; |
449 |
uint32_t limit; |
uint32_t limit; |
450 |
|
|
451 |
/* |
/* There is no MBTrans8to16 for Inter block, that's done in motion compensation |
452 |
* There is no MBTrans8to16 for Inter block, that's done in motion compensation |
* already */ |
|
* already |
|
|
*/ |
|
453 |
|
|
454 |
/* Perform DCT (and field decision) */ |
/* Perform DCT (and field decision) */ |
455 |
MBfDCT(pParam, frame, pMB, x_pos, y_pos, data); |
MBfDCT(pParam, frame, pMB, x_pos, y_pos, data); |
487 |
uint8_t cbp; |
uint8_t cbp; |
488 |
uint32_t limit; |
uint32_t limit; |
489 |
|
|
490 |
/* |
/* There is no MBTrans8to16 for Inter block, that's done in motion compensation |
491 |
* There is no MBTrans8to16 for Inter block, that's done in motion compensation |
* already */ |
|
* already |
|
|
*/ |
|
492 |
|
|
493 |
/* Perform DCT (and field decision) */ |
/* Perform DCT (and field decision) */ |
494 |
MBfDCT(pParam, frame, pMB, x_pos, y_pos, data); |
MBfDCT(pParam, frame, pMB, x_pos, y_pos, data); |
506 |
* History comment: |
* History comment: |
507 |
* We don't have to DeQuant, iDCT and Transfer back data for B-frames. |
* We don't have to DeQuant, iDCT and Transfer back data for B-frames. |
508 |
* |
* |
509 |
* BUT some plugins require the original frame to be passed so we have |
* BUT some plugins require the rebuilt original frame to be passed so we |
510 |
* to take care of that here |
* have to take care of that here |
511 |
*/ |
*/ |
512 |
if((pParam->plugin_flags & XVID_REQORIGINAL)) { |
if((pParam->plugin_flags & XVID_REQORIGINAL)) { |
513 |
|
|
632 |
MOVLINE(LINE(3, 3), tmp); |
MOVLINE(LINE(3, 3), tmp); |
633 |
} |
} |
634 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
635 |
/***************************************************************************** |
/***************************************************************************** |
636 |
* Trellis based R-D optimal quantization |
* Trellis based R-D optimal quantization |
637 |
* |
* |
639 |
* |
* |
640 |
****************************************************************************/ |
****************************************************************************/ |
641 |
|
|
|
|
|
|
#if 0 |
|
|
static int |
|
|
dct_quantize_trellis_mpeg_c(int16_t *const Out, |
|
|
const int16_t *const In, |
|
|
int Q, |
|
|
const uint16_t * const Zigzag, |
|
|
int Non_Zero) |
|
|
{ |
|
|
return 63; |
|
|
} |
|
|
#endif |
|
|
|
|
642 |
/*---------------------------------------------------------------------------- |
/*---------------------------------------------------------------------------- |
643 |
* |
* |
644 |
* Trellis-Based quantization |
* Trellis-Based quantization |
650 |
* IEEE Transactions on Image Processing, Vol.9, No.8, Aug. 2000. |
* IEEE Transactions on Image Processing, Vol.9, No.8, Aug. 2000. |
651 |
* |
* |
652 |
* we are at stake with a simplified Bellmand-Ford / Dijkstra Single |
* we are at stake with a simplified Bellmand-Ford / Dijkstra Single |
653 |
* Source Shorted Path algo. But due to the underlying graph structure |
* Source Shortest Path algo. But due to the underlying graph structure |
654 |
* ("Trellis"), it can be turned into a dynamic programming algo, |
* ("Trellis"), it can be turned into a dynamic programming algo, |
655 |
* partially saving the explicit graph's nodes representation. And |
* partially saving the explicit graph's nodes representation. And |
656 |
* without using a heap, since the open frontier of the DAG is always |
* without using a heap, since the open frontier of the DAG is always |
657 |
* known, and of fixed sized. |
* known, and of fixed size. |
658 |
*--------------------------------------------------------------------------*/ |
*--------------------------------------------------------------------------*/ |
659 |
|
|
660 |
|
|
754 |
Code_Len24,Code_Len23,Code_Len22,Code_Len21, Code_Len3, Code_Len1, |
Code_Len24,Code_Len23,Code_Len22,Code_Len21, Code_Len3, Code_Len1, |
755 |
}; |
}; |
756 |
|
|
757 |
#define TL(q) 0xfe00/(q*q) |
/* TL_SHIFT controls the precision of the RD optimizations in trellis |
758 |
|
* valid range is [10..16]. The bigger, the more trellis is vulnerable |
759 |
|
* to overflows in cost formulas. |
760 |
|
* - 10 allows ac values up to 2^11 == 2048 |
761 |
|
* - 16 allows ac values up to 2^8 == 256 |
762 |
|
*/ |
763 |
|
#define TL_SHIFT 11 |
764 |
|
#define TL(q) ((0xfe00>>(16-TL_SHIFT))/(q*q)) |
765 |
|
|
766 |
static const int Trellis_Lambda_Tabs[31] = { |
static const int Trellis_Lambda_Tabs[31] = { |
767 |
TL( 1),TL( 2),TL( 3),TL( 4),TL( 5),TL( 6), TL( 7), |
TL( 1),TL( 2),TL( 3),TL( 4),TL( 5),TL( 6), TL( 7), |
791 |
|
|
792 |
return(sum); |
return(sum); |
793 |
} |
} |
|
/* this routine has been strippen of all debug code */ |
|
794 |
|
|
795 |
|
/* this routine has been strippen of all debug code */ |
796 |
static int |
static int |
797 |
dct_quantize_trellis_h263_c(int16_t *const Out, const int16_t *const In, int Q, const uint16_t * const Zigzag, int Non_Zero) |
dct_quantize_trellis_c(int16_t *const Out, |
798 |
|
const int16_t *const In, |
799 |
|
int Q, |
800 |
|
const uint16_t * const Zigzag, |
801 |
|
const uint16_t * const QuantMatrix, |
802 |
|
int Non_Zero) |
803 |
{ |
{ |
804 |
|
|
805 |
/* |
/* Note: We should search last non-zero coeffs on *real* DCT input coeffs |
806 |
* Note: We should search last non-zero coeffs on *real* DCT input coeffs (In[]), |
* (In[]), not quantized one (Out[]). However, it only improves the result |
807 |
* not quantized one (Out[]). However, it only improves the result *very* |
* *very* slightly (~0.01dB), whereas speed drops to crawling level :) |
808 |
* slightly (~0.01dB), whereas speed drops to crawling level :) |
* Well, actually, taking 1 more coeff past Non_Zero into account sometimes |
809 |
* Well, actually, taking 1 more coeff past Non_Zero into account sometimes helps. |
* helps. */ |
|
*/ |
|
810 |
typedef struct { int16_t Run, Level; } NODE; |
typedef struct { int16_t Run, Level; } NODE; |
811 |
|
|
812 |
NODE Nodes[65], Last; |
NODE Nodes[65], Last; |
813 |
uint32_t Run_Costs0[64+1]; |
uint32_t Run_Costs0[64+1]; |
814 |
uint32_t * const Run_Costs = Run_Costs0 + 1; |
uint32_t * const Run_Costs = Run_Costs0 + 1; |
815 |
const int Mult = 2*Q; |
|
816 |
const int Bias = (Q-1) | 1; |
/* it's 1/lambda, actually */ |
817 |
const int Lev0 = Mult + Bias; |
const int Lambda = Trellis_Lambda_Tabs[Q-1]; |
|
const int Lambda = Trellis_Lambda_Tabs[Q-1]; /* it's 1/lambda, actually */ |
|
818 |
|
|
819 |
int Run_Start = -1; |
int Run_Start = -1; |
820 |
uint32_t Min_Cost = 2<<16; |
uint32_t Min_Cost = 2<<TL_SHIFT; |
821 |
|
|
822 |
int Last_Node = -1; |
int Last_Node = -1; |
823 |
uint32_t Last_Cost = 0; |
uint32_t Last_Cost = 0; |
824 |
|
|
825 |
int i, j, sum; |
int i, j, sum; |
826 |
Run_Costs[-1] = 2<<16; /* source (w/ CBP penalty) */ |
|
827 |
|
/* source (w/ CBP penalty) */ |
828 |
|
Run_Costs[-1] = 2<<TL_SHIFT; |
829 |
|
|
830 |
Non_Zero = Find_Last(Out, Zigzag, Non_Zero); |
Non_Zero = Find_Last(Out, Zigzag, Non_Zero); |
831 |
if (Non_Zero<0) |
if (Non_Zero<0) |
832 |
return 0; /* Sum is zero if there are only zero coeffs */ |
return 0; /* Sum is zero if there are only zero coeffs */ |
833 |
|
|
834 |
for(i=0; i<=Non_Zero; i++) { |
for(i=0; i<=Non_Zero; i++) { |
835 |
|
const int q = ((Q*QuantMatrix[Zigzag[i]])>>4); |
836 |
|
const int Mult = 2*q; |
837 |
|
const int Bias = (q-1) | 1; |
838 |
|
const int Lev0 = Mult + Bias; |
839 |
|
|
840 |
const int AC = In[Zigzag[i]]; |
const int AC = In[Zigzag[i]]; |
841 |
const int Level1 = Out[Zigzag[i]]; |
const int Level1 = Out[Zigzag[i]]; |
842 |
const int Dist0 = Lambda* AC*AC; |
const unsigned int Dist0 = Lambda* AC*AC; |
843 |
uint32_t Best_Cost = 0xf0000000; |
uint32_t Best_Cost = 0xf0000000; |
844 |
Last_Cost += Dist0; |
Last_Cost += Dist0; |
845 |
|
|
859 |
Cost0 = Lambda*dQ*dQ; |
Cost0 = Lambda*dQ*dQ; |
860 |
|
|
861 |
Nodes[i].Run = 1; |
Nodes[i].Run = 1; |
862 |
Best_Cost = (Code_Len20[0]<<16) + Run_Costs[i-1]+Cost0; |
Best_Cost = (Code_Len20[0]<<TL_SHIFT) + Run_Costs[i-1]+Cost0; |
863 |
for(Run=i-Run_Start; Run>0; --Run) { |
for(Run=i-Run_Start; Run>0; --Run) { |
864 |
const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run]; |
const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run]; |
865 |
const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<16); |
const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<TL_SHIFT); |
866 |
const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<16); |
const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<TL_SHIFT); |
867 |
|
|
868 |
/* |
/* TODO: what about tie-breaks? Should we favor short runs or |
|
* TODO: what about tie-breaks? Should we favor short runs or |
|
869 |
* long runs? Although the error is the same, it would not be |
* long runs? Although the error is the same, it would not be |
870 |
* spread the same way along high and low frequencies... |
* spread the same way along high and low frequencies... */ |
|
*/ |
|
871 |
|
|
872 |
/* (I'd say: favour short runs => hifreq errors (HVS) -- gruel ) */ |
/* Gruel: I'd say, favour short runs => hifreq errors (HVS) */ |
873 |
|
|
874 |
if (Cost<Best_Cost) { |
if (Cost<Best_Cost) { |
875 |
Best_Cost = Cost; |
Best_Cost = Cost; |
884 |
} |
} |
885 |
if (Last_Node==i) |
if (Last_Node==i) |
886 |
Last.Level = Nodes[i].Level; |
Last.Level = Nodes[i].Level; |
887 |
} else { /* "big" levels */ |
} else if (51U>(uint32_t)(Level1+25)) { |
888 |
|
/* "big" levels (not less than ESC3, though) */ |
889 |
const uint8_t *Tbl_L1, *Tbl_L2, *Tbl_L1_Last, *Tbl_L2_Last; |
const uint8_t *Tbl_L1, *Tbl_L2, *Tbl_L1_Last, *Tbl_L2_Last; |
890 |
int Level2; |
int Level2; |
891 |
int dQ1, dQ2; |
int dQ1, dQ2; |
921 |
uint32_t Cost1, Cost2; |
uint32_t Cost1, Cost2; |
922 |
int bLevel; |
int bLevel; |
923 |
|
|
924 |
/* |
/* for sub-optimal (but slightly worth it, speed-wise) search, |
925 |
* for sub-optimal (but slightly worth it, speed-wise) search, uncomment the following: |
* uncomment the following: |
926 |
* if (Cost_Base>=Best_Cost) continue; |
* if (Cost_Base>=Best_Cost) continue; |
927 |
* (? doesn't seem to have any effect -- gruel ) |
* (? doesn't seem to have any effect -- gruel ) */ |
|
*/ |
|
928 |
|
|
929 |
Cost1 = Cost_Base + (Tbl_L1[Run-1]<<16); |
Cost1 = Cost_Base + (Tbl_L1[Run-1]<<TL_SHIFT); |
930 |
Cost2 = Cost_Base + (Tbl_L2[Run-1]<<16) + dDist21; |
Cost2 = Cost_Base + (Tbl_L2[Run-1]<<TL_SHIFT) + dDist21; |
931 |
|
|
932 |
if (Cost2<Cost1) { |
if (Cost2<Cost1) { |
933 |
Cost1 = Cost2; |
Cost1 = Cost2; |
942 |
Nodes[i].Level = bLevel; |
Nodes[i].Level = bLevel; |
943 |
} |
} |
944 |
|
|
945 |
Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<16); |
Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<TL_SHIFT); |
946 |
Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<16) + dDist21; |
Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<TL_SHIFT) + dDist21; |
947 |
|
|
948 |
if (Cost2<Cost1) { |
if (Cost2<Cost1) { |
949 |
Cost1 = Cost2; |
Cost1 = Cost2; |
959 |
Last_Node = i; |
Last_Node = i; |
960 |
} |
} |
961 |
} /* end of "for Run" */ |
} /* end of "for Run" */ |
962 |
|
} else { |
963 |
|
/* Very very high levels, with no chance of being optimizable |
964 |
|
* => Simply pick best Run. */ |
965 |
|
int Run; |
966 |
|
for(Run=i-Run_Start; Run>0; --Run) { |
967 |
|
/* 30 bits + no distortion */ |
968 |
|
const uint32_t Cost = (30<<TL_SHIFT) + Run_Costs[i-Run]; |
969 |
|
if (Cost<Best_Cost) { |
970 |
|
Best_Cost = Cost; |
971 |
|
Nodes[i].Run = Run; |
972 |
|
Nodes[i].Level = Level1; |
973 |
|
} |
974 |
|
|
975 |
|
if (Cost<Last_Cost) { |
976 |
|
Last_Cost = Cost; |
977 |
|
Last.Run = Run; |
978 |
|
Last.Level = Level1; |
979 |
|
Last_Node = i; |
980 |
|
} |
981 |
} |
} |
982 |
|
} |
983 |
|
|
984 |
|
|
985 |
Run_Costs[i] = Best_Cost; |
Run_Costs[i] = Best_Cost; |
986 |
|
|
988 |
Min_Cost = Best_Cost; |
Min_Cost = Best_Cost; |
989 |
Run_Start = i; |
Run_Start = i; |
990 |
} else { |
} else { |
991 |
/* |
/* as noticed by Michael Niedermayer (michaelni at gmx.at), |
992 |
* as noticed by Michael Niedermayer (michaelni at gmx.at), there's |
* there's a code shorter by 1 bit for a larger run (!), same |
993 |
* a code shorter by 1 bit for a larger run (!), same level. We give |
* level. We give it a chance by not moving the left barrier too |
994 |
* it a chance by not moving the left barrier too much. |
* much. */ |
995 |
*/ |
while( Run_Costs[Run_Start]>Min_Cost+(1<<TL_SHIFT) ) |
|
|
|
|
while( Run_Costs[Run_Start]>Min_Cost+(1<<16) ) |
|
996 |
Run_Start++; |
Run_Start++; |
997 |
|
|
998 |
/* spread on preceding coeffs the cost incurred by skipping this one */ |
/* spread on preceding coeffs the cost incurred by skipping this |
999 |
|
* one */ |
1000 |
for(j=Run_Start; j<i; ++j) Run_Costs[j] += Dist0; |
for(j=Run_Start; j<i; ++j) Run_Costs[j] += Dist0; |
1001 |
Min_Cost += Dist0; |
Min_Cost += Dist0; |
1002 |
} |
} |
1003 |
} |
} |
1004 |
|
|
1005 |
/* It seems trellis doesn't give good results... just compute the Out sum and |
/* It seems trellis doesn't give good results... just compute the Out sum |
1006 |
* quit (even if we did not modify it, upperlayer relies on this data) */ |
* and quit */ |
1007 |
if (Last_Node<0) |
if (Last_Node<0) |
1008 |
return Compute_Sum(Out, Non_Zero); |
return Compute_Sum(Out, Non_Zero); |
1009 |
|
|
1021 |
return sum; |
return sum; |
1022 |
} |
} |
1023 |
|
|
|
static int |
|
|
dct_quantize_trellis_mpeg_c(int16_t *const Out, const int16_t *const In, int Q, const uint16_t * const Zigzag, int Non_Zero) |
|
|
{ |
|
|
/* ToDo: Ok ok it's just a place holder for Gruel -- damn write this one :-) */ |
|
|
return Compute_Sum(Out, 63); |
|
|
} |
|
|
|
|
1024 |
/* original version including heavy debugging info */ |
/* original version including heavy debugging info */ |
1025 |
|
|
1026 |
#ifdef DBGTRELL |
#ifdef DBGTRELL |
1077 |
V -= Ref[Zigzag[i]]; |
V -= Ref[Zigzag[i]]; |
1078 |
Dist += V*V; |
Dist += V*V; |
1079 |
} |
} |
1080 |
Cost = Lambda*Dist + (Bits<<16); |
Cost = Lambda*Dist + (Bits<<TL_SHIFT); |
1081 |
if (DBG==1) |
if (DBG==1) |
1082 |
printf( " Last:%2d/%2d Cost = [(Bits=%5.0d) + Lambda*(Dist=%6.0d) = %d ] >>12= %d ", Last,Max, Bits, Dist, Cost, Cost>>12 ); |
printf( " Last:%2d/%2d Cost = [(Bits=%5.0d) + Lambda*(Dist=%6.0d) = %d ] >>12= %d ", Last,Max, Bits, Dist, Cost, Cost>>12 ); |
1083 |
return Cost; |
return Cost; |
1109 |
const int Lambda = Trellis_Lambda_Tabs[Q-1]; /* it's 1/lambda, actually */ |
const int Lambda = Trellis_Lambda_Tabs[Q-1]; /* it's 1/lambda, actually */ |
1110 |
|
|
1111 |
int Run_Start = -1; |
int Run_Start = -1; |
1112 |
Run_Costs[-1] = 2<<16; /* source (w/ CBP penalty) */ |
Run_Costs[-1] = 2<<TL_SHIFT; /* source (w/ CBP penalty) */ |
1113 |
uint32_t Min_Cost = 2<<16; |
uint32_t Min_Cost = 2<<TL_SHIFT; |
1114 |
|
|
1115 |
int Last_Node = -1; |
int Last_Node = -1; |
1116 |
uint32_t Last_Cost = 0; |
uint32_t Last_Cost = 0; |
1149 |
Cost0 = Lambda*dQ*dQ; |
Cost0 = Lambda*dQ*dQ; |
1150 |
|
|
1151 |
Nodes[i].Run = 1; |
Nodes[i].Run = 1; |
1152 |
Best_Cost = (Code_Len20[0]<<16) + Run_Costs[i-1]+Cost0; |
Best_Cost = (Code_Len20[0]<<TL_SHIFT) + Run_Costs[i-1]+Cost0; |
1153 |
for(Run=i-Run_Start; Run>0; --Run) |
for(Run=i-Run_Start; Run>0; --Run) |
1154 |
{ |
{ |
1155 |
const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run]; |
const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run]; |
1156 |
const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<16); |
const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<TL_SHIFT); |
1157 |
const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<16); |
const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<TL_SHIFT); |
1158 |
|
|
1159 |
/* |
/* |
1160 |
* TODO: what about tie-breaks? Should we favor short runs or |
* TODO: what about tie-breaks? Should we favor short runs or |
1230 |
* for sub-optimal (but slightly worth it, speed-wise) search, uncomment the following: |
* for sub-optimal (but slightly worth it, speed-wise) search, uncomment the following: |
1231 |
* if (Cost_Base>=Best_Cost) continue; |
* if (Cost_Base>=Best_Cost) continue; |
1232 |
*/ |
*/ |
1233 |
Cost1 = Cost_Base + (Tbl_L1[Run-1]<<16); |
Cost1 = Cost_Base + (Tbl_L1[Run-1]<<TL_SHIFT); |
1234 |
Cost2 = Cost_Base + (Tbl_L2[Run-1]<<16) + dDist21; |
Cost2 = Cost_Base + (Tbl_L2[Run-1]<<TL_SHIFT) + dDist21; |
1235 |
|
|
1236 |
if (Cost2<Cost1) { |
if (Cost2<Cost1) { |
1237 |
Cost1 = Cost2; |
Cost1 = Cost2; |
1245 |
Nodes[i].Level = bLevel; |
Nodes[i].Level = bLevel; |
1246 |
} |
} |
1247 |
|
|
1248 |
Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<16); |
Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<TL_SHIFT); |
1249 |
Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<16) + dDist21; |
Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<TL_SHIFT) + dDist21; |
1250 |
|
|
1251 |
if (Cost2<Cost1) { |
if (Cost2<Cost1) { |
1252 |
Cost1 = Cost2; |
Cost1 = Cost2; |
1292 |
* it a chance by not moving the left barrier too much. |
* it a chance by not moving the left barrier too much. |
1293 |
*/ |
*/ |
1294 |
|
|
1295 |
while( Run_Costs[Run_Start]>Min_Cost+(1<<16) ) |
while( Run_Costs[Run_Start]>Min_Cost+(1<<TL_SHIFT) ) |
1296 |
Run_Start++; |
Run_Start++; |
1297 |
|
|
1298 |
/* spread on preceding coeffs the cost incurred by skipping this one */ |
/* spread on preceding coeffs the cost incurred by skipping this one */ |