[cvs] / xvidcore / src / utils / mbtransquant.c Repository:
ViewVC logotype

Annotation of /xvidcore/src/utils/mbtransquant.c

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.31 - (view) (download)

1 : edgomez 1.23 /*****************************************************************************
2 :     *
3 :     * XVID MPEG-4 VIDEO CODEC
4 :     * - MB Transfert/Quantization functions -
5 :     *
6 :     * Copyright(C) 2001-2003 Peter Ross <pross@xvid.org>
7 :     * 2001-2003 Michael Militzer <isibaar@xvid.org>
8 :     * 2003 Edouard Gomez <ed.gomez@free.fr>
9 :     *
10 :     * This program is free software ; you can redistribute it and/or modify
11 :     * it under the terms of the GNU General Public License as published by
12 :     * the Free Software Foundation ; either version 2 of the License, or
13 :     * (at your option) any later version.
14 :     *
15 :     * This program is distributed in the hope that it will be useful,
16 :     * but WITHOUT ANY WARRANTY ; without even the implied warranty of
17 :     * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 :     * GNU General Public License for more details.
19 :     *
20 :     * You should have received a copy of the GNU General Public License
21 :     * along with this program ; if not, write to the Free Software
22 :     * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 :     *
24 : syskin 1.31 * $Id: mbtransquant.c,v 1.30 2005/12/09 04:45:35 syskin Exp $
25 : edgomez 1.23 *
26 :     ****************************************************************************/
27 : Isibaar 1.1
28 : edgomez 1.23 #include <stdio.h>
29 :     #include <stdlib.h>
30 : edgomez 1.3 #include <string.h>
31 :    
32 : Isibaar 1.1 #include "../portab.h"
33 :     #include "mbfunctions.h"
34 :    
35 :     #include "../global.h"
36 :     #include "mem_transfer.h"
37 :     #include "timer.h"
38 : edgomez 1.23 #include "../bitstream/mbcoding.h"
39 :     #include "../bitstream/zigzag.h"
40 : Isibaar 1.1 #include "../dct/fdct.h"
41 :     #include "../dct/idct.h"
42 : edgomez 1.23 #include "../quant/quant.h"
43 : Isibaar 1.1 #include "../encoder.h"
44 :    
45 : edgomez 1.23 #include "../quant/quant_matrix.h"
46 : Isibaar 1.1
47 : edgomez 1.21 MBFIELDTEST_PTR MBFieldTest;
48 : Isibaar 1.1
49 : edgomez 1.23 /*
50 :     * Skip blocks having a coefficient sum below this value. This value will be
51 :     * corrected according to the MB quantizer to avoid artifacts for quant==1
52 :     */
53 :     #define PVOP_TOOSMALL_LIMIT 1
54 :     #define BVOP_TOOSMALL_LIMIT 3
55 :    
56 :     /*****************************************************************************
57 :     * Local functions
58 :     ****************************************************************************/
59 : Isibaar 1.1
60 : edgomez 1.23 /* permute block and return field dct choice */
61 :     static __inline uint32_t
62 :     MBDecideFieldDCT(int16_t data[6 * 64])
63 :     {
64 :     uint32_t field = MBFieldTest(data);
65 :    
66 :     if (field)
67 :     MBFrameToField(data);
68 :    
69 :     return field;
70 :     }
71 :    
72 :     /* Performs Forward DCT on all blocks */
73 : syskin 1.22 static __inline void
74 : edgomez 1.23 MBfDCT(const MBParam * const pParam,
75 :     const FRAMEINFO * const frame,
76 :     MACROBLOCK * const pMB,
77 :     uint32_t x_pos,
78 :     uint32_t y_pos,
79 :     int16_t data[6 * 64])
80 : syskin 1.22 {
81 : edgomez 1.23 /* Handles interlacing */
82 :     start_timer();
83 :     pMB->field_dct = 0;
84 :     if ((frame->vol_flags & XVID_VOL_INTERLACING) &&
85 :     (x_pos>0) && (x_pos<pParam->mb_width-1) &&
86 :     (y_pos>0) && (y_pos<pParam->mb_height-1)) {
87 :     pMB->field_dct = MBDecideFieldDCT(data);
88 :     }
89 :     stop_interlacing_timer();
90 :    
91 :     /* Perform DCT */
92 : syskin 1.22 start_timer();
93 : suxen_drol 1.29 fdct((short * const)&data[0 * 64]);
94 :     fdct((short * const)&data[1 * 64]);
95 :     fdct((short * const)&data[2 * 64]);
96 :     fdct((short * const)&data[3 * 64]);
97 :     fdct((short * const)&data[4 * 64]);
98 :     fdct((short * const)&data[5 * 64]);
99 : syskin 1.22 stop_dct_timer();
100 :     }
101 :    
102 : edgomez 1.23 /* Performs Inverse DCT on all blocks */
103 :     static __inline void
104 :     MBiDCT(int16_t data[6 * 64],
105 :     const uint8_t cbp)
106 :     {
107 :     start_timer();
108 : suxen_drol 1.29 if(cbp & (1 << (5 - 0))) idct((short * const)&data[0 * 64]);
109 :     if(cbp & (1 << (5 - 1))) idct((short * const)&data[1 * 64]);
110 :     if(cbp & (1 << (5 - 2))) idct((short * const)&data[2 * 64]);
111 :     if(cbp & (1 << (5 - 3))) idct((short * const)&data[3 * 64]);
112 :     if(cbp & (1 << (5 - 4))) idct((short * const)&data[4 * 64]);
113 :     if(cbp & (1 << (5 - 5))) idct((short * const)&data[5 * 64]);
114 : edgomez 1.23 stop_idct_timer();
115 :     }
116 : syskin 1.22
117 : edgomez 1.23 /* Quantize all blocks -- Intra mode */
118 :     static __inline void
119 :     MBQuantIntra(const MBParam * pParam,
120 :     const FRAMEINFO * const frame,
121 :     const MACROBLOCK * pMB,
122 :     int16_t qcoeff[6 * 64],
123 :     int16_t data[6*64])
124 : syskin 1.22 {
125 : edgomez 1.23 int mpeg;
126 :     int scaler_lum, scaler_chr;
127 :    
128 :     quant_intraFuncPtr const quant[2] =
129 :     {
130 :     quant_h263_intra,
131 :     quant_mpeg_intra
132 :     };
133 :    
134 :     mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
135 :     scaler_lum = get_dc_scaler(pMB->quant, 1);
136 :     scaler_chr = get_dc_scaler(pMB->quant, 0);
137 : syskin 1.22
138 : edgomez 1.23 /* Quantize the block */
139 : syskin 1.22 start_timer();
140 : edgomez 1.23 quant[mpeg](&data[0 * 64], &qcoeff[0 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices);
141 :     quant[mpeg](&data[1 * 64], &qcoeff[1 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices);
142 :     quant[mpeg](&data[2 * 64], &qcoeff[2 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices);
143 :     quant[mpeg](&data[3 * 64], &qcoeff[3 * 64], pMB->quant, scaler_lum, pParam->mpeg_quant_matrices);
144 :     quant[mpeg](&data[4 * 64], &qcoeff[4 * 64], pMB->quant, scaler_chr, pParam->mpeg_quant_matrices);
145 :     quant[mpeg](&data[5 * 64], &qcoeff[5 * 64], pMB->quant, scaler_chr, pParam->mpeg_quant_matrices);
146 : syskin 1.22 stop_quant_timer();
147 :     }
148 :    
149 : edgomez 1.23 /* DeQuantize all blocks -- Intra mode */
150 :     static __inline void
151 :     MBDeQuantIntra(const MBParam * pParam,
152 :     const int iQuant,
153 :     int16_t qcoeff[6 * 64],
154 :     int16_t data[6*64])
155 : Isibaar 1.1 {
156 : edgomez 1.23 int mpeg;
157 :     int scaler_lum, scaler_chr;
158 : edgomez 1.3
159 : edgomez 1.23 quant_intraFuncPtr const dequant[2] =
160 :     {
161 :     dequant_h263_intra,
162 :     dequant_mpeg_intra
163 :     };
164 :    
165 :     mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
166 :     scaler_lum = get_dc_scaler(iQuant, 1);
167 :     scaler_chr = get_dc_scaler(iQuant, 0);
168 : Isibaar 1.1
169 : h 1.2 start_timer();
170 : edgomez 1.23 dequant[mpeg](&qcoeff[0 * 64], &data[0 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices);
171 :     dequant[mpeg](&qcoeff[1 * 64], &data[1 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices);
172 :     dequant[mpeg](&qcoeff[2 * 64], &data[2 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices);
173 :     dequant[mpeg](&qcoeff[3 * 64], &data[3 * 64], iQuant, scaler_lum, pParam->mpeg_quant_matrices);
174 :     dequant[mpeg](&qcoeff[4 * 64], &data[4 * 64], iQuant, scaler_chr, pParam->mpeg_quant_matrices);
175 :     dequant[mpeg](&qcoeff[5 * 64], &data[5 * 64], iQuant, scaler_chr, pParam->mpeg_quant_matrices);
176 :     stop_iquant_timer();
177 :     }
178 :    
179 :     static int
180 :     dct_quantize_trellis_c(int16_t *const Out,
181 :     const int16_t *const In,
182 :     int Q,
183 :     const uint16_t * const Zigzag,
184 :     const uint16_t * const QuantMatrix,
185 : edgomez 1.24 int Non_Zero,
186 : syskin 1.30 int Sum,
187 :     int Lambda_Mod);
188 : edgomez 1.23
189 :     /* Quantize all blocks -- Inter mode */
190 :     static __inline uint8_t
191 :     MBQuantInter(const MBParam * pParam,
192 :     const FRAMEINFO * const frame,
193 :     const MACROBLOCK * pMB,
194 :     int16_t data[6 * 64],
195 :     int16_t qcoeff[6 * 64],
196 :     int bvop,
197 :     int limit)
198 :     {
199 :    
200 :     int i;
201 :     uint8_t cbp = 0;
202 :     int sum;
203 :     int code_block, mpeg;
204 : h 1.2
205 : edgomez 1.23 quant_interFuncPtr const quant[2] =
206 :     {
207 :     quant_h263_inter,
208 :     quant_mpeg_inter
209 :     };
210 : h 1.2
211 : edgomez 1.23 mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
212 : syskin 1.22
213 : edgomez 1.7 for (i = 0; i < 6; i++) {
214 : Isibaar 1.1
215 : edgomez 1.23 /* Quantize the block */
216 : Isibaar 1.1 start_timer();
217 : edgomez 1.23
218 :     sum = quant[mpeg](&qcoeff[i*64], &data[i*64], pMB->quant, pParam->mpeg_quant_matrices);
219 :    
220 : edgomez 1.27 if(sum && (pMB->quant > 2) && (frame->vop_flags & XVID_VOP_TRELLISQUANT)) {
221 : edgomez 1.23 const uint16_t *matrix;
222 :     const static uint16_t h263matrix[] =
223 :     {
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 :     16, 16, 16, 16, 16, 16, 16, 16
232 :     };
233 :    
234 :     matrix = (mpeg)?get_inter_matrix(pParam->mpeg_quant_matrices):h263matrix;
235 :     sum = dct_quantize_trellis_c(&qcoeff[i*64], &data[i*64],
236 :     pMB->quant, &scan_tables[0][0],
237 :     matrix,
238 : edgomez 1.24 63,
239 : syskin 1.30 sum,
240 :     pMB->lambda[i]);
241 : edgomez 1.23 }
242 : syskin 1.22 stop_quant_timer();
243 : edgomez 1.21
244 : edgomez 1.23 /*
245 :     * We code the block if the sum is higher than the limit and if the first
246 :     * two AC coefficients in zig zag order are not zero.
247 :     */
248 :     code_block = 0;
249 :     if ((sum >= limit) || (qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) {
250 :     code_block = 1;
251 :     } else {
252 : Isibaar 1.1
253 : edgomez 1.23 if (bvop && (pMB->mode == MODE_DIRECT || pMB->mode == MODE_DIRECT_NO4V)) {
254 :     /* dark blocks prevention for direct mode */
255 :     if ((qcoeff[i*64] < -1) || (qcoeff[i*64] > 0))
256 :     code_block = 1;
257 :     } else {
258 :     /* not direct mode */
259 :     if (qcoeff[i*64] != 0)
260 :     code_block = 1;
261 :     }
262 : edgomez 1.21 }
263 : edgomez 1.23
264 :     /* Set the corresponding cbp bit */
265 :     cbp |= code_block << (5 - i);
266 : edgomez 1.21 }
267 :    
268 : edgomez 1.23 return(cbp);
269 :     }
270 :    
271 :     /* DeQuantize all blocks -- Inter mode */
272 :     static __inline void
273 :     MBDeQuantInter(const MBParam * pParam,
274 :     const int iQuant,
275 :     int16_t data[6 * 64],
276 :     int16_t qcoeff[6 * 64],
277 :     const uint8_t cbp)
278 :     {
279 :     int mpeg;
280 : edgomez 1.21
281 : edgomez 1.23 quant_interFuncPtr const dequant[2] =
282 :     {
283 :     dequant_h263_inter,
284 :     dequant_mpeg_inter
285 :     };
286 : Isibaar 1.1
287 : edgomez 1.23 mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
288 : Isibaar 1.1
289 : edgomez 1.23 start_timer();
290 :     if(cbp & (1 << (5 - 0))) dequant[mpeg](&data[0 * 64], &qcoeff[0 * 64], iQuant, pParam->mpeg_quant_matrices);
291 :     if(cbp & (1 << (5 - 1))) dequant[mpeg](&data[1 * 64], &qcoeff[1 * 64], iQuant, pParam->mpeg_quant_matrices);
292 :     if(cbp & (1 << (5 - 2))) dequant[mpeg](&data[2 * 64], &qcoeff[2 * 64], iQuant, pParam->mpeg_quant_matrices);
293 :     if(cbp & (1 << (5 - 3))) dequant[mpeg](&data[3 * 64], &qcoeff[3 * 64], iQuant, pParam->mpeg_quant_matrices);
294 :     if(cbp & (1 << (5 - 4))) dequant[mpeg](&data[4 * 64], &qcoeff[4 * 64], iQuant, pParam->mpeg_quant_matrices);
295 :     if(cbp & (1 << (5 - 5))) dequant[mpeg](&data[5 * 64], &qcoeff[5 * 64], iQuant, pParam->mpeg_quant_matrices);
296 :     stop_iquant_timer();
297 : Isibaar 1.1 }
298 :    
299 : edgomez 1.23 typedef void (transfer_operation_8to16_t) (int16_t *Dst, const uint8_t *Src, int BpS);
300 :     typedef void (transfer_operation_16to8_t) (uint8_t *Dst, const int16_t *Src, int BpS);
301 :    
302 :    
303 :     static __inline void
304 :     MBTrans8to16(const MBParam * const pParam,
305 :     const FRAMEINFO * const frame,
306 :     const MACROBLOCK * const pMB,
307 :     const uint32_t x_pos,
308 :     const uint32_t y_pos,
309 :     int16_t data[6 * 64])
310 : Isibaar 1.1 {
311 : h 1.4 uint32_t stride = pParam->edged_width;
312 : edgomez 1.23 uint32_t stride2 = stride / 2;
313 :     uint32_t next_block = stride * 8;
314 : Isibaar 1.1 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
315 : syskin 1.22 const IMAGE * const pCurrent = &frame->image;
316 :    
317 : edgomez 1.23 /* Image pointers */
318 : syskin 1.26 pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4);
319 :     pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3);
320 :     pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3);
321 : Isibaar 1.1
322 : edgomez 1.23 /* Do the transfer */
323 :     start_timer();
324 : syskin 1.26 transfer_8to16copy(&data[0 * 64], pY_Cur, stride);
325 :     transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride);
326 :     transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride);
327 :     transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride);
328 :     transfer_8to16copy(&data[4 * 64], pU_Cur, stride2);
329 :     transfer_8to16copy(&data[5 * 64], pV_Cur, stride2);
330 : edgomez 1.23 stop_transfer_timer();
331 :     }
332 : Isibaar 1.1
333 : edgomez 1.23 static __inline void
334 :     MBTrans16to8(const MBParam * const pParam,
335 :     const FRAMEINFO * const frame,
336 :     const MACROBLOCK * const pMB,
337 :     const uint32_t x_pos,
338 :     const uint32_t y_pos,
339 :     int16_t data[6 * 64],
340 :     const uint32_t add, /* Must be 1 or 0 */
341 :     const uint8_t cbp)
342 :     {
343 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
344 :     uint32_t stride = pParam->edged_width;
345 :     uint32_t stride2 = stride / 2;
346 :     uint32_t next_block = stride * 8;
347 :     const IMAGE * const pCurrent = &frame->image;
348 : Isibaar 1.1
349 : syskin 1.26 /* Array of function pointers, indexed by [add] */
350 :     transfer_operation_16to8_t * const functions[2] =
351 : edgomez 1.23 {
352 :     (transfer_operation_16to8_t*)transfer_16to8copy,
353 :     (transfer_operation_16to8_t*)transfer_16to8add,
354 :     };
355 :    
356 :     transfer_operation_16to8_t *transfer_op = NULL;
357 :    
358 :     /* Image pointers */
359 : syskin 1.26 pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4);
360 :     pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3);
361 :     pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3);
362 : h 1.2
363 : edgomez 1.7 if (pMB->field_dct) {
364 : h 1.4 next_block = stride;
365 :     stride *= 2;
366 : h 1.2 }
367 :    
368 : edgomez 1.23 /* Operation function */
369 : syskin 1.26 transfer_op = functions[add];
370 : edgomez 1.23
371 :     /* Do the operation */
372 : h 1.2 start_timer();
373 : syskin 1.26 if (cbp&32) transfer_op(pY_Cur, &data[0 * 64], stride);
374 :     if (cbp&16) transfer_op(pY_Cur + 8, &data[1 * 64], stride);
375 :     if (cbp& 8) transfer_op(pY_Cur + next_block, &data[2 * 64], stride);
376 :     if (cbp& 4) transfer_op(pY_Cur + next_block + 8, &data[3 * 64], stride);
377 :     if (cbp& 2) transfer_op(pU_Cur, &data[4 * 64], stride2);
378 :     if (cbp& 1) transfer_op(pV_Cur, &data[5 * 64], stride2);
379 : h 1.2 stop_transfer_timer();
380 : edgomez 1.23 }
381 :    
382 :     /*****************************************************************************
383 :     * Module functions
384 :     ****************************************************************************/
385 :    
386 :     void
387 :     MBTransQuantIntra(const MBParam * const pParam,
388 :     const FRAMEINFO * const frame,
389 :     MACROBLOCK * const pMB,
390 :     const uint32_t x_pos,
391 :     const uint32_t y_pos,
392 :     int16_t data[6 * 64],
393 :     int16_t qcoeff[6 * 64])
394 :     {
395 :    
396 :     /* Transfer data */
397 :     MBTrans8to16(pParam, frame, pMB, x_pos, y_pos, data);
398 :    
399 :     /* Perform DCT (and field decision) */
400 :     MBfDCT(pParam, frame, pMB, x_pos, y_pos, data);
401 :    
402 :     /* Quantize the block */
403 :     MBQuantIntra(pParam, frame, pMB, data, qcoeff);
404 : h 1.2
405 : edgomez 1.23 /* DeQuantize the block */
406 :     MBDeQuantIntra(pParam, pMB->quant, data, qcoeff);
407 :    
408 :     /* Perform inverse DCT*/
409 :     MBiDCT(data, 0x3F);
410 :    
411 :     /* Transfer back the data -- Don't add data */
412 :     MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 0, 0x3F);
413 : chl 1.8 }
414 :    
415 : edgomez 1.23
416 : chl 1.8 uint8_t
417 : edgomez 1.23 MBTransQuantInter(const MBParam * const pParam,
418 :     const FRAMEINFO * const frame,
419 :     MACROBLOCK * const pMB,
420 :     const uint32_t x_pos,
421 :     const uint32_t y_pos,
422 : chl 1.8 int16_t data[6 * 64],
423 :     int16_t qcoeff[6 * 64])
424 :     {
425 : edgomez 1.23 uint8_t cbp;
426 :     uint32_t limit;
427 :    
428 :     /* There is no MBTrans8to16 for Inter block, that's done in motion compensation
429 :     * already */
430 :    
431 :     /* Perform DCT (and field decision) */
432 :     MBfDCT(pParam, frame, pMB, x_pos, y_pos, data);
433 :    
434 :     /* Set the limit threshold */
435 :     limit = PVOP_TOOSMALL_LIMIT + ((pMB->quant == 1)? 1 : 0);
436 :    
437 :     if (frame->vop_flags & XVID_VOP_CARTOON)
438 :     limit *= 3;
439 :    
440 :     /* Quantize the block */
441 :     cbp = MBQuantInter(pParam, frame, pMB, data, qcoeff, 0, limit);
442 : chl 1.8
443 : edgomez 1.23 /* DeQuantize the block */
444 :     MBDeQuantInter(pParam, pMB->quant, data, qcoeff, cbp);
445 : chl 1.8
446 : edgomez 1.23 /* Perform inverse DCT*/
447 :     MBiDCT(data, cbp);
448 : chl 1.8
449 : edgomez 1.23 /* Transfer back the data -- Add the data */
450 :     MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 1, cbp);
451 : chl 1.8
452 : edgomez 1.23 return(cbp);
453 : chl 1.8 }
454 :    
455 : edgomez 1.23 uint8_t
456 :     MBTransQuantInterBVOP(const MBParam * pParam,
457 :     FRAMEINFO * frame,
458 :     MACROBLOCK * pMB,
459 :     const uint32_t x_pos,
460 :     const uint32_t y_pos,
461 :     int16_t data[6 * 64],
462 :     int16_t qcoeff[6 * 64])
463 :     {
464 :     uint8_t cbp;
465 :     uint32_t limit;
466 :    
467 :     /* There is no MBTrans8to16 for Inter block, that's done in motion compensation
468 :     * already */
469 :    
470 :     /* Perform DCT (and field decision) */
471 :     MBfDCT(pParam, frame, pMB, x_pos, y_pos, data);
472 :    
473 :     /* Set the limit threshold */
474 :     limit = BVOP_TOOSMALL_LIMIT;
475 :    
476 :     if (frame->vop_flags & XVID_VOP_CARTOON)
477 :     limit *= 2;
478 :    
479 :     /* Quantize the block */
480 :     cbp = MBQuantInter(pParam, frame, pMB, data, qcoeff, 1, limit);
481 :    
482 :     /*
483 :     * History comment:
484 :     * We don't have to DeQuant, iDCT and Transfer back data for B-frames.
485 :     *
486 :     * BUT some plugins require the rebuilt original frame to be passed so we
487 :     * have to take care of that here
488 :     */
489 :     if((pParam->plugin_flags & XVID_REQORIGINAL)) {
490 :    
491 :     /* DeQuantize the block */
492 :     MBDeQuantInter(pParam, pMB->quant, data, qcoeff, cbp);
493 :    
494 :     /* Perform inverse DCT*/
495 :     MBiDCT(data, cbp);
496 : h 1.2
497 : edgomez 1.23 /* Transfer back the data -- Add the data */
498 :     MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 1, cbp);
499 :     }
500 : edgomez 1.21
501 : edgomez 1.23 return(cbp);
502 : edgomez 1.21 }
503 : edgomez 1.3
504 : edgomez 1.21 /* if sum(diff between field lines) < sum(diff between frame lines), use field dct */
505 :     uint32_t
506 :     MBFieldTest_c(int16_t data[6 * 64])
507 :     {
508 : edgomez 1.7 const uint8_t blocks[] =
509 :     { 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 };
510 :     const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 };
511 : h 1.2
512 :     int frame = 0, field = 0;
513 :     int i, j;
514 :    
515 : edgomez 1.7 for (i = 0; i < 7; ++i) {
516 :     for (j = 0; j < 8; ++j) {
517 :     frame +=
518 : edgomez 1.23 abs(data[0 * 64 + (i + 1) * 8 + j] - data[0 * 64 + i * 8 + j]);
519 : edgomez 1.7 frame +=
520 : edgomez 1.23 abs(data[1 * 64 + (i + 1) * 8 + j] - data[1 * 64 + i * 8 + j]);
521 : edgomez 1.7 frame +=
522 : edgomez 1.23 abs(data[2 * 64 + (i + 1) * 8 + j] - data[2 * 64 + i * 8 + j]);
523 : edgomez 1.7 frame +=
524 : edgomez 1.23 abs(data[3 * 64 + (i + 1) * 8 + j] - data[3 * 64 + i * 8 + j]);
525 : edgomez 1.7
526 :     field +=
527 : edgomez 1.23 abs(data[blocks[i + 1] + lines[i + 1] + j] -
528 : edgomez 1.7 data[blocks[i] + lines[i] + j]);
529 :     field +=
530 : edgomez 1.23 abs(data[blocks[i + 1] + lines[i + 1] + 8 + j] -
531 : edgomez 1.7 data[blocks[i] + lines[i] + 8 + j]);
532 :     field +=
533 : edgomez 1.23 abs(data[blocks[i + 1] + 64 + lines[i + 1] + j] -
534 : edgomez 1.7 data[blocks[i] + 64 + lines[i] + j]);
535 :     field +=
536 : edgomez 1.23 abs(data[blocks[i + 1] + 64 + lines[i + 1] + 8 + j] -
537 : edgomez 1.7 data[blocks[i] + 64 + lines[i] + 8 + j]);
538 : Isibaar 1.1 }
539 :     }
540 : h 1.2
541 : edgomez 1.21 return (frame >= (field + 350));
542 : h 1.2 }
543 :    
544 :    
545 :     /* deinterlace Y blocks vertically */
546 :    
547 :     #define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp))
548 : edgomez 1.23 #define LINE(X,Y) &data[X*64 + Y*8]
549 : h 1.2
550 : edgomez 1.7 void
551 :     MBFrameToField(int16_t data[6 * 64])
552 : h 1.2 {
553 :     int16_t tmp[8];
554 :    
555 :     /* left blocks */
556 :    
557 : edgomez 1.23 /* 1=2, 2=4, 4=8, 8=1 */
558 : edgomez 1.7 MOVLINE(tmp, LINE(0, 1));
559 :     MOVLINE(LINE(0, 1), LINE(0, 2));
560 :     MOVLINE(LINE(0, 2), LINE(0, 4));
561 :     MOVLINE(LINE(0, 4), LINE(2, 0));
562 :     MOVLINE(LINE(2, 0), tmp);
563 : h 1.2
564 : edgomez 1.23 /* 3=6, 6=12, 12=9, 9=3 */
565 : edgomez 1.7 MOVLINE(tmp, LINE(0, 3));
566 :     MOVLINE(LINE(0, 3), LINE(0, 6));
567 :     MOVLINE(LINE(0, 6), LINE(2, 4));
568 :     MOVLINE(LINE(2, 4), LINE(2, 1));
569 :     MOVLINE(LINE(2, 1), tmp);
570 : h 1.2
571 : edgomez 1.23 /* 5=10, 10=5 */
572 : edgomez 1.7 MOVLINE(tmp, LINE(0, 5));
573 :     MOVLINE(LINE(0, 5), LINE(2, 2));
574 :     MOVLINE(LINE(2, 2), tmp);
575 : h 1.2
576 : edgomez 1.23 /* 7=14, 14=13, 13=11, 11=7 */
577 : edgomez 1.7 MOVLINE(tmp, LINE(0, 7));
578 :     MOVLINE(LINE(0, 7), LINE(2, 6));
579 :     MOVLINE(LINE(2, 6), LINE(2, 5));
580 :     MOVLINE(LINE(2, 5), LINE(2, 3));
581 :     MOVLINE(LINE(2, 3), tmp);
582 : h 1.2
583 :     /* right blocks */
584 :    
585 : edgomez 1.23 /* 1=2, 2=4, 4=8, 8=1 */
586 : edgomez 1.7 MOVLINE(tmp, LINE(1, 1));
587 :     MOVLINE(LINE(1, 1), LINE(1, 2));
588 :     MOVLINE(LINE(1, 2), LINE(1, 4));
589 :     MOVLINE(LINE(1, 4), LINE(3, 0));
590 :     MOVLINE(LINE(3, 0), tmp);
591 : h 1.2
592 : edgomez 1.23 /* 3=6, 6=12, 12=9, 9=3 */
593 : edgomez 1.7 MOVLINE(tmp, LINE(1, 3));
594 :     MOVLINE(LINE(1, 3), LINE(1, 6));
595 :     MOVLINE(LINE(1, 6), LINE(3, 4));
596 :     MOVLINE(LINE(3, 4), LINE(3, 1));
597 :     MOVLINE(LINE(3, 1), tmp);
598 : h 1.2
599 : edgomez 1.23 /* 5=10, 10=5 */
600 : edgomez 1.7 MOVLINE(tmp, LINE(1, 5));
601 :     MOVLINE(LINE(1, 5), LINE(3, 2));
602 :     MOVLINE(LINE(3, 2), tmp);
603 : h 1.2
604 : edgomez 1.23 /* 7=14, 14=13, 13=11, 11=7 */
605 : edgomez 1.7 MOVLINE(tmp, LINE(1, 7));
606 :     MOVLINE(LINE(1, 7), LINE(3, 6));
607 :     MOVLINE(LINE(3, 6), LINE(3, 5));
608 :     MOVLINE(LINE(3, 5), LINE(3, 3));
609 :     MOVLINE(LINE(3, 3), tmp);
610 : Isibaar 1.1 }
611 : edgomez 1.23
612 :     /*****************************************************************************
613 :     * Trellis based R-D optimal quantization
614 :     *
615 :     * Trellis Quant code (C) 2003 Pascal Massimino skal(at)planet-d.net
616 :     *
617 :     ****************************************************************************/
618 :    
619 :     /*----------------------------------------------------------------------------
620 :     *
621 :     * Trellis-Based quantization
622 :     *
623 :     * So far I understand this paper:
624 :     *
625 :     * "Trellis-Based R-D Optimal Quantization in H.263+"
626 :     * J.Wen, M.Luttrell, J.Villasenor
627 :     * IEEE Transactions on Image Processing, Vol.9, No.8, Aug. 2000.
628 :     *
629 :     * we are at stake with a simplified Bellmand-Ford / Dijkstra Single
630 :     * Source Shortest Path algo. But due to the underlying graph structure
631 :     * ("Trellis"), it can be turned into a dynamic programming algo,
632 :     * partially saving the explicit graph's nodes representation. And
633 :     * without using a heap, since the open frontier of the DAG is always
634 :     * known, and of fixed size.
635 :     *--------------------------------------------------------------------------*/
636 :    
637 :    
638 :    
639 :     /* Codes lengths for relevant levels. */
640 :    
641 :     /* let's factorize: */
642 :     static const uint8_t Code_Len0[64] = {
643 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
644 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
645 :     static const uint8_t Code_Len1[64] = {
646 :     20,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
647 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
648 :     static const uint8_t Code_Len2[64] = {
649 :     19,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
650 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
651 :     static const uint8_t Code_Len3[64] = {
652 :     18,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
653 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
654 :     static const uint8_t Code_Len4[64] = {
655 :     17,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
656 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
657 :     static const uint8_t Code_Len5[64] = {
658 :     16,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
659 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
660 :     static const uint8_t Code_Len6[64] = {
661 :     15,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
662 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
663 :     static const uint8_t Code_Len7[64] = {
664 :     13,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
665 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
666 :     static const uint8_t Code_Len8[64] = {
667 :     11,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
668 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
669 :     static const uint8_t Code_Len9[64] = {
670 :     12,21,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
671 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
672 :     static const uint8_t Code_Len10[64] = {
673 :     12,20,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
674 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
675 :     static const uint8_t Code_Len11[64] = {
676 :     12,19,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
677 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
678 :     static const uint8_t Code_Len12[64] = {
679 :     11,17,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
680 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
681 :     static const uint8_t Code_Len13[64] = {
682 :     11,15,21,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
683 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
684 :     static const uint8_t Code_Len14[64] = {
685 :     10,12,19,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
686 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
687 :     static const uint8_t Code_Len15[64] = {
688 :     10,13,17,19,21,21,21,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
689 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
690 :     static const uint8_t Code_Len16[64] = {
691 :     9,12,13,18,18,19,19,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
692 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30};
693 :     static const uint8_t Code_Len17[64] = {
694 :     8,11,13,14,14,14,15,19,19,19,21,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
695 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
696 :     static const uint8_t Code_Len18[64] = {
697 :     7, 9,11,11,13,13,13,15,15,15,16,22,22,22,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
698 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
699 :     static const uint8_t Code_Len19[64] = {
700 :     5, 7, 9,10,10,11,11,11,11,11,13,14,16,17,17,18,18,18,18,18,18,18,18,20,20,21,21,30,30,30,30,30,
701 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
702 :     static const uint8_t Code_Len20[64] = {
703 :     3, 4, 5, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8, 9, 9,10,10,10,10,10,10,10,10,12,12,13,13,12,13,14,15,15,
704 :     15,16,16,16,16,17,17,17,18,18,19,19,19,19,19,19,19,19,21,21,22,22,30,30,30,30,30,30,30,30,30,30 };
705 :    
706 :     /* a few more table for LAST table: */
707 :     static const uint8_t Code_Len21[64] = {
708 :     13,20,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
709 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30};
710 :     static const uint8_t Code_Len22[64] = {
711 :     12,15,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,
712 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30};
713 :     static const uint8_t Code_Len23[64] = {
714 :     10,12,15,15,15,16,16,16,16,17,17,17,17,17,17,17,17,18,18,18,18,18,18,18,18,19,19,19,19,20,20,20,
715 :     20,21,21,21,21,21,21,21,21,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30};
716 :     static const uint8_t Code_Len24[64] = {
717 :     5, 7, 7, 7, 7, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9,10,10,10,10,10,10,10,10,11,11,11,11,12,12,12,
718 :     12,13,13,13,13,13,13,13,13,14,16,16,16,16,17,17,17,17,18,18,18,18,18,18,18,18,19,19,19,19,19,19};
719 :    
720 :    
721 :     static const uint8_t * const B16_17_Code_Len[24] = { /* levels [1..24] */
722 :     Code_Len20,Code_Len19,Code_Len18,Code_Len17,
723 :     Code_Len16,Code_Len15,Code_Len14,Code_Len13,
724 :     Code_Len12,Code_Len11,Code_Len10,Code_Len9,
725 :     Code_Len8, Code_Len7 ,Code_Len6 ,Code_Len5,
726 :     Code_Len4, Code_Len3, Code_Len3 ,Code_Len2,
727 :     Code_Len2, Code_Len1, Code_Len1, Code_Len1,
728 :     };
729 :    
730 :     static const uint8_t * const B16_17_Code_Len_Last[6] = { /* levels [1..6] */
731 :     Code_Len24,Code_Len23,Code_Len22,Code_Len21, Code_Len3, Code_Len1,
732 :     };
733 :    
734 :     /* TL_SHIFT controls the precision of the RD optimizations in trellis
735 :     * valid range is [10..16]. The bigger, the more trellis is vulnerable
736 :     * to overflows in cost formulas.
737 :     * - 10 allows ac values up to 2^11 == 2048
738 :     * - 16 allows ac values up to 2^8 == 256
739 :     */
740 :     #define TL_SHIFT 11
741 :     #define TL(q) ((0xfe00>>(16-TL_SHIFT))/(q*q))
742 :    
743 :     static const int Trellis_Lambda_Tabs[31] = {
744 :     TL( 1),TL( 2),TL( 3),TL( 4),TL( 5),TL( 6), TL( 7),
745 :     TL( 8),TL( 9),TL(10),TL(11),TL(12),TL(13),TL(14), TL(15),
746 :     TL(16),TL(17),TL(18),TL(19),TL(20),TL(21),TL(22), TL(23),
747 :     TL(24),TL(25),TL(26),TL(27),TL(28),TL(29),TL(30), TL(31)
748 :     };
749 :     #undef TL
750 :    
751 :     static int __inline
752 :     Find_Last(const int16_t *C, const uint16_t *Zigzag, int i)
753 :     {
754 :     while(i>=0)
755 :     if (C[Zigzag[i]])
756 :     return i;
757 :     else i--;
758 :     return -1;
759 :     }
760 :    
761 : syskin 1.31 #define TRELLIS_MIN_EFFORT 3
762 :    
763 : edgomez 1.23 /* this routine has been strippen of all debug code */
764 :     static int
765 :     dct_quantize_trellis_c(int16_t *const Out,
766 :     const int16_t *const In,
767 :     int Q,
768 :     const uint16_t * const Zigzag,
769 :     const uint16_t * const QuantMatrix,
770 : edgomez 1.24 int Non_Zero,
771 : syskin 1.30 int Sum,
772 :     int Lambda_Mod)
773 : edgomez 1.23 {
774 :    
775 :     /* Note: We should search last non-zero coeffs on *real* DCT input coeffs
776 :     * (In[]), not quantized one (Out[]). However, it only improves the result
777 :     * *very* slightly (~0.01dB), whereas speed drops to crawling level :)
778 :     * Well, actually, taking 1 more coeff past Non_Zero into account sometimes
779 :     * helps. */
780 :     typedef struct { int16_t Run, Level; } NODE;
781 :    
782 : Skal 1.28 NODE Nodes[65], Last = { 0, 0};
783 : edgomez 1.23 uint32_t Run_Costs0[64+1];
784 :     uint32_t * const Run_Costs = Run_Costs0 + 1;
785 :    
786 :     /* it's 1/lambda, actually */
787 : syskin 1.30 const int Lambda = (Lambda_Mod*Trellis_Lambda_Tabs[Q-1])>>LAMBDA_EXP;
788 : edgomez 1.23
789 :     int Run_Start = -1;
790 :     uint32_t Min_Cost = 2<<TL_SHIFT;
791 :    
792 :     int Last_Node = -1;
793 :     uint32_t Last_Cost = 0;
794 :    
795 : edgomez 1.24 int i, j;
796 : edgomez 1.23
797 :     /* source (w/ CBP penalty) */
798 :     Run_Costs[-1] = 2<<TL_SHIFT;
799 :    
800 :     Non_Zero = Find_Last(Out, Zigzag, Non_Zero);
801 : syskin 1.31 if (Non_Zero < TRELLIS_MIN_EFFORT)
802 :     Non_Zero = TRELLIS_MIN_EFFORT;
803 : edgomez 1.23
804 :     for(i=0; i<=Non_Zero; i++) {
805 :     const int q = ((Q*QuantMatrix[Zigzag[i]])>>4);
806 :     const int Mult = 2*q;
807 :     const int Bias = (q-1) | 1;
808 :     const int Lev0 = Mult + Bias;
809 :    
810 :     const int AC = In[Zigzag[i]];
811 :     const int Level1 = Out[Zigzag[i]];
812 :     const unsigned int Dist0 = Lambda* AC*AC;
813 :     uint32_t Best_Cost = 0xf0000000;
814 :     Last_Cost += Dist0;
815 :    
816 :     /* very specialized loop for -1,0,+1 */
817 :     if ((uint32_t)(Level1+1)<3) {
818 :     int dQ;
819 :     int Run;
820 :     uint32_t Cost0;
821 :    
822 :     if (AC<0) {
823 :     Nodes[i].Level = -1;
824 :     dQ = Lev0 + AC;
825 :     } else {
826 :     Nodes[i].Level = 1;
827 :     dQ = Lev0 - AC;
828 :     }
829 :     Cost0 = Lambda*dQ*dQ;
830 :    
831 :     Nodes[i].Run = 1;
832 :     Best_Cost = (Code_Len20[0]<<TL_SHIFT) + Run_Costs[i-1]+Cost0;
833 :     for(Run=i-Run_Start; Run>0; --Run) {
834 :     const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run];
835 :     const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<TL_SHIFT);
836 :     const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<TL_SHIFT);
837 :    
838 :     /* TODO: what about tie-breaks? Should we favor short runs or
839 :     * long runs? Although the error is the same, it would not be
840 :     * spread the same way along high and low frequencies... */
841 :    
842 :     /* Gruel: I'd say, favour short runs => hifreq errors (HVS) */
843 :    
844 :     if (Cost<Best_Cost) {
845 :     Best_Cost = Cost;
846 :     Nodes[i].Run = Run;
847 :     }
848 :    
849 :     if (lCost<Last_Cost) {
850 :     Last_Cost = lCost;
851 :     Last.Run = Run;
852 :     Last_Node = i;
853 :     }
854 :     }
855 :     if (Last_Node==i)
856 :     Last.Level = Nodes[i].Level;
857 :     } else if (51U>(uint32_t)(Level1+25)) {
858 :     /* "big" levels (not less than ESC3, though) */
859 :     const uint8_t *Tbl_L1, *Tbl_L2, *Tbl_L1_Last, *Tbl_L2_Last;
860 :     int Level2;
861 :     int dQ1, dQ2;
862 :     int Run;
863 :     uint32_t Dist1,Dist2;
864 :     int dDist21;
865 :    
866 :     if (Level1>1) {
867 :     dQ1 = Level1*Mult-AC + Bias;
868 :     dQ2 = dQ1 - Mult;
869 :     Level2 = Level1-1;
870 :     Tbl_L1 = (Level1<=24) ? B16_17_Code_Len[Level1-1] : Code_Len0;
871 :     Tbl_L2 = (Level2<=24) ? B16_17_Code_Len[Level2-1] : Code_Len0;
872 :     Tbl_L1_Last = (Level1<=6) ? B16_17_Code_Len_Last[Level1-1] : Code_Len0;
873 :     Tbl_L2_Last = (Level2<=6) ? B16_17_Code_Len_Last[Level2-1] : Code_Len0;
874 :     } else { /* Level1<-1 */
875 :     dQ1 = Level1*Mult-AC - Bias;
876 :     dQ2 = dQ1 + Mult;
877 :     Level2 = Level1 + 1;
878 :     Tbl_L1 = (Level1>=-24) ? B16_17_Code_Len[Level1^-1] : Code_Len0;
879 :     Tbl_L2 = (Level2>=-24) ? B16_17_Code_Len[Level2^-1] : Code_Len0;
880 :     Tbl_L1_Last = (Level1>=- 6) ? B16_17_Code_Len_Last[Level1^-1] : Code_Len0;
881 :     Tbl_L2_Last = (Level2>=- 6) ? B16_17_Code_Len_Last[Level2^-1] : Code_Len0;
882 :     }
883 :    
884 :     Dist1 = Lambda*dQ1*dQ1;
885 :     Dist2 = Lambda*dQ2*dQ2;
886 :     dDist21 = Dist2-Dist1;
887 :    
888 :     for(Run=i-Run_Start; Run>0; --Run)
889 :     {
890 :     const uint32_t Cost_Base = Dist1 + Run_Costs[i-Run];
891 :     uint32_t Cost1, Cost2;
892 :     int bLevel;
893 :    
894 :     /* for sub-optimal (but slightly worth it, speed-wise) search,
895 :     * uncomment the following:
896 :     * if (Cost_Base>=Best_Cost) continue;
897 :     * (? doesn't seem to have any effect -- gruel ) */
898 :    
899 :     Cost1 = Cost_Base + (Tbl_L1[Run-1]<<TL_SHIFT);
900 :     Cost2 = Cost_Base + (Tbl_L2[Run-1]<<TL_SHIFT) + dDist21;
901 :    
902 :     if (Cost2<Cost1) {
903 :     Cost1 = Cost2;
904 :     bLevel = Level2;
905 :     } else {
906 :     bLevel = Level1;
907 :     }
908 :    
909 :     if (Cost1<Best_Cost) {
910 :     Best_Cost = Cost1;
911 :     Nodes[i].Run = Run;
912 :     Nodes[i].Level = bLevel;
913 :     }
914 :    
915 :     Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<TL_SHIFT);
916 :     Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<TL_SHIFT) + dDist21;
917 :    
918 :     if (Cost2<Cost1) {
919 :     Cost1 = Cost2;
920 :     bLevel = Level2;
921 :     } else {
922 :     bLevel = Level1;
923 :     }
924 :    
925 :     if (Cost1<Last_Cost) {
926 :     Last_Cost = Cost1;
927 :     Last.Run = Run;
928 :     Last.Level = bLevel;
929 :     Last_Node = i;
930 :     }
931 :     } /* end of "for Run" */
932 :     } else {
933 :     /* Very very high levels, with no chance of being optimizable
934 :     * => Simply pick best Run. */
935 :     int Run;
936 :     for(Run=i-Run_Start; Run>0; --Run) {
937 :     /* 30 bits + no distortion */
938 :     const uint32_t Cost = (30<<TL_SHIFT) + Run_Costs[i-Run];
939 :     if (Cost<Best_Cost) {
940 :     Best_Cost = Cost;
941 :     Nodes[i].Run = Run;
942 :     Nodes[i].Level = Level1;
943 :     }
944 :    
945 :     if (Cost<Last_Cost) {
946 :     Last_Cost = Cost;
947 :     Last.Run = Run;
948 :     Last.Level = Level1;
949 :     Last_Node = i;
950 :     }
951 :     }
952 :     }
953 :    
954 :    
955 :     Run_Costs[i] = Best_Cost;
956 :    
957 :     if (Best_Cost < Min_Cost + Dist0) {
958 :     Min_Cost = Best_Cost;
959 :     Run_Start = i;
960 :     } else {
961 :     /* as noticed by Michael Niedermayer (michaelni at gmx.at),
962 :     * there's a code shorter by 1 bit for a larger run (!), same
963 :     * level. We give it a chance by not moving the left barrier too
964 :     * much. */
965 :     while( Run_Costs[Run_Start]>Min_Cost+(1<<TL_SHIFT) )
966 :     Run_Start++;
967 :    
968 :     /* spread on preceding coeffs the cost incurred by skipping this
969 :     * one */
970 :     for(j=Run_Start; j<i; ++j) Run_Costs[j] += Dist0;
971 :     Min_Cost += Dist0;
972 :     }
973 :     }
974 :    
975 : edgomez 1.24 /* It seems trellis doesn't give good results... just leave the block untouched
976 :     * and return the original sum value */
977 : edgomez 1.23 if (Last_Node<0)
978 : edgomez 1.24 return Sum;
979 : edgomez 1.23
980 :     /* reconstruct optimal sequence backward with surviving paths */
981 :     memset(Out, 0x00, 64*sizeof(*Out));
982 :     Out[Zigzag[Last_Node]] = Last.Level;
983 :     i = Last_Node - Last.Run;
984 : syskin 1.25 Sum = abs(Last.Level);
985 : edgomez 1.23 while(i>=0) {
986 :     Out[Zigzag[i]] = Nodes[i].Level;
987 : edgomez 1.24 Sum += abs(Nodes[i].Level);
988 : edgomez 1.23 i -= Nodes[i].Run;
989 :     }
990 :    
991 : edgomez 1.24 return Sum;
992 : edgomez 1.23 }
993 :    
994 :     /* original version including heavy debugging info */
995 :    
996 :     #ifdef DBGTRELL
997 :    
998 :     #define DBG 0
999 :    
1000 :     static __inline uint32_t Evaluate_Cost(const int16_t *C, int Mult, int Bias,
1001 :     const uint16_t * Zigzag, int Max, int Lambda)
1002 :     {
1003 :     #if (DBG>0)
1004 :     const int16_t * const Ref = C + 6*64;
1005 :     int Last = Max;
1006 :     int Bits = 0;
1007 :     int Dist = 0;
1008 :     int i;
1009 :     uint32_t Cost;
1010 :    
1011 :     while(Last>=0 && C[Zigzag[Last]]==0)
1012 :     Last--;
1013 :    
1014 :     if (Last>=0) {
1015 :     int j=0, j0=0;
1016 :     int Run, Level;
1017 :    
1018 :     Bits = 2; /* CBP */
1019 :     while(j<Last) {
1020 :     while(!C[Zigzag[j]])
1021 :     j++;
1022 :     if (j==Last)
1023 :     break;
1024 :     Level=C[Zigzag[j]];
1025 :     Run = j - j0;
1026 :     j0 = ++j;
1027 :     if (Level>=-24 && Level<=24)
1028 :     Bits += B16_17_Code_Len[(Level<0) ? -Level-1 : Level-1][Run];
1029 :     else
1030 :     Bits += 30;
1031 :     }
1032 :     Level = C[Zigzag[Last]];
1033 :     Run = j - j0;
1034 :     if (Level>=-6 && Level<=6)
1035 :     Bits += B16_17_Code_Len_Last[(Level<0) ? -Level-1 : Level-1][Run];
1036 :     else
1037 :     Bits += 30;
1038 :     }
1039 :    
1040 :     for(i=0; i<=Last; ++i) {
1041 :     int V = C[Zigzag[i]]*Mult;
1042 :     if (V>0)
1043 :     V += Bias;
1044 :     else
1045 :     if (V<0)
1046 :     V -= Bias;
1047 :     V -= Ref[Zigzag[i]];
1048 :     Dist += V*V;
1049 :     }
1050 :     Cost = Lambda*Dist + (Bits<<TL_SHIFT);
1051 :     if (DBG==1)
1052 :     printf( " Last:%2d/%2d Cost = [(Bits=%5.0d) + Lambda*(Dist=%6.0d) = %d ] >>12= %d ", Last,Max, Bits, Dist, Cost, Cost>>12 );
1053 :     return Cost;
1054 :    
1055 :     #else
1056 :     return 0;
1057 :     #endif
1058 :     }
1059 :    
1060 :    
1061 :     static int
1062 :     dct_quantize_trellis_h263_c(int16_t *const Out, const int16_t *const In, int Q, const uint16_t * const Zigzag, int Non_Zero)
1063 :     {
1064 :    
1065 :     /*
1066 :     * Note: We should search last non-zero coeffs on *real* DCT input coeffs (In[]),
1067 :     * not quantized one (Out[]). However, it only improves the result *very*
1068 :     * slightly (~0.01dB), whereas speed drops to crawling level :)
1069 :     * Well, actually, taking 1 more coeff past Non_Zero into account sometimes helps.
1070 :     */
1071 :     typedef struct { int16_t Run, Level; } NODE;
1072 :    
1073 :     NODE Nodes[65], Last;
1074 :     uint32_t Run_Costs0[64+1];
1075 :     uint32_t * const Run_Costs = Run_Costs0 + 1;
1076 :     const int Mult = 2*Q;
1077 :     const int Bias = (Q-1) | 1;
1078 :     const int Lev0 = Mult + Bias;
1079 :     const int Lambda = Trellis_Lambda_Tabs[Q-1]; /* it's 1/lambda, actually */
1080 :    
1081 :     int Run_Start = -1;
1082 :     Run_Costs[-1] = 2<<TL_SHIFT; /* source (w/ CBP penalty) */
1083 :     uint32_t Min_Cost = 2<<TL_SHIFT;
1084 :    
1085 :     int Last_Node = -1;
1086 :     uint32_t Last_Cost = 0;
1087 :    
1088 :     int i, j;
1089 :    
1090 :     #if (DBG>0)
1091 :     Last.Level = 0; Last.Run = -1; /* just initialize to smthg */
1092 :     #endif
1093 :    
1094 :     Non_Zero = Find_Last(Out, Zigzag, Non_Zero);
1095 :     if (Non_Zero<0)
1096 :     return -1;
1097 :    
1098 :     for(i=0; i<=Non_Zero; i++)
1099 :     {
1100 :     const int AC = In[Zigzag[i]];
1101 :     const int Level1 = Out[Zigzag[i]];
1102 :     const int Dist0 = Lambda* AC*AC;
1103 :     uint32_t Best_Cost = 0xf0000000;
1104 :     Last_Cost += Dist0;
1105 :    
1106 :     if ((uint32_t)(Level1+1)<3) /* very specialized loop for -1,0,+1 */
1107 :     {
1108 :     int dQ;
1109 :     int Run;
1110 :     uint32_t Cost0;
1111 :    
1112 :     if (AC<0) {
1113 :     Nodes[i].Level = -1;
1114 :     dQ = Lev0 + AC;
1115 :     } else {
1116 :     Nodes[i].Level = 1;
1117 :     dQ = Lev0 - AC;
1118 :     }
1119 :     Cost0 = Lambda*dQ*dQ;
1120 :    
1121 :     Nodes[i].Run = 1;
1122 :     Best_Cost = (Code_Len20[0]<<TL_SHIFT) + Run_Costs[i-1]+Cost0;
1123 :     for(Run=i-Run_Start; Run>0; --Run)
1124 :     {
1125 :     const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run];
1126 :     const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<TL_SHIFT);
1127 :     const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<TL_SHIFT);
1128 :    
1129 :     /*
1130 :     * TODO: what about tie-breaks? Should we favor short runs or
1131 :     * long runs? Although the error is the same, it would not be
1132 :     * spread the same way along high and low frequencies...
1133 :     */
1134 :     if (Cost<Best_Cost) {
1135 :     Best_Cost = Cost;
1136 :     Nodes[i].Run = Run;
1137 :     }
1138 :    
1139 :     if (lCost<Last_Cost) {
1140 :     Last_Cost = lCost;
1141 :     Last.Run = Run;
1142 :     Last_Node = i;
1143 :     }
1144 :     }
1145 :     if (Last_Node==i)
1146 :     Last.Level = Nodes[i].Level;
1147 :    
1148 :     if (DBG==1) {
1149 :     Run_Costs[i] = Best_Cost;
1150 :     printf( "Costs #%2d: ", i);
1151 :     for(j=-1;j<=Non_Zero;++j) {
1152 :     if (j==Run_Start) printf( " %3.0d|", Run_Costs[j]>>12 );
1153 :     else if (j>Run_Start && j<i) printf( " %3.0d|", Run_Costs[j]>>12 );
1154 :     else if (j==i) printf( "(%3.0d)", Run_Costs[j]>>12 );
1155 :     else printf( " - |" );
1156 :     }
1157 :     printf( "<%3.0d %2d %d>", Min_Cost>>12, Nodes[i].Level, Nodes[i].Run );
1158 :     printf( " Last:#%2d {%3.0d %2d %d}", Last_Node, Last_Cost>>12, Last.Level, Last.Run );
1159 :     printf( " AC:%3.0d Dist0:%3d Dist(%d)=%d", AC, Dist0>>12, Nodes[i].Level, Cost0>>12 );
1160 :     printf( "\n" );
1161 :     }
1162 :     }
1163 :     else /* "big" levels */
1164 :     {
1165 :     const uint8_t *Tbl_L1, *Tbl_L2, *Tbl_L1_Last, *Tbl_L2_Last;
1166 :     int Level2;
1167 :     int dQ1, dQ2;
1168 :     int Run;
1169 :     uint32_t Dist1,Dist2;
1170 :     int dDist21;
1171 :    
1172 :     if (Level1>1) {
1173 :     dQ1 = Level1*Mult-AC + Bias;
1174 :     dQ2 = dQ1 - Mult;
1175 :     Level2 = Level1-1;
1176 :     Tbl_L1 = (Level1<=24) ? B16_17_Code_Len[Level1-1] : Code_Len0;
1177 :     Tbl_L2 = (Level2<=24) ? B16_17_Code_Len[Level2-1] : Code_Len0;
1178 :     Tbl_L1_Last = (Level1<=6) ? B16_17_Code_Len_Last[Level1-1] : Code_Len0;
1179 :     Tbl_L2_Last = (Level2<=6) ? B16_17_Code_Len_Last[Level2-1] : Code_Len0;
1180 :     } else { /* Level1<-1 */
1181 :     dQ1 = Level1*Mult-AC - Bias;
1182 :     dQ2 = dQ1 + Mult;
1183 :     Level2 = Level1 + 1;
1184 :     Tbl_L1 = (Level1>=-24) ? B16_17_Code_Len[Level1^-1] : Code_Len0;
1185 :     Tbl_L2 = (Level2>=-24) ? B16_17_Code_Len[Level2^-1] : Code_Len0;
1186 :     Tbl_L1_Last = (Level1>=- 6) ? B16_17_Code_Len_Last[Level1^-1] : Code_Len0;
1187 :     Tbl_L2_Last = (Level2>=- 6) ? B16_17_Code_Len_Last[Level2^-1] : Code_Len0;
1188 :     }
1189 :     Dist1 = Lambda*dQ1*dQ1;
1190 :     Dist2 = Lambda*dQ2*dQ2;
1191 :     dDist21 = Dist2-Dist1;
1192 :    
1193 :     for(Run=i-Run_Start; Run>0; --Run)
1194 :     {
1195 :     const uint32_t Cost_Base = Dist1 + Run_Costs[i-Run];
1196 :     uint32_t Cost1, Cost2;
1197 :     int bLevel;
1198 :    
1199 :     /*
1200 :     * for sub-optimal (but slightly worth it, speed-wise) search, uncomment the following:
1201 :     * if (Cost_Base>=Best_Cost) continue;
1202 :     */
1203 :     Cost1 = Cost_Base + (Tbl_L1[Run-1]<<TL_SHIFT);
1204 :     Cost2 = Cost_Base + (Tbl_L2[Run-1]<<TL_SHIFT) + dDist21;
1205 :    
1206 :     if (Cost2<Cost1) {
1207 :     Cost1 = Cost2;
1208 :     bLevel = Level2;
1209 :     } else
1210 :     bLevel = Level1;
1211 :    
1212 :     if (Cost1<Best_Cost) {
1213 :     Best_Cost = Cost1;
1214 :     Nodes[i].Run = Run;
1215 :     Nodes[i].Level = bLevel;
1216 :     }
1217 :    
1218 :     Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<TL_SHIFT);
1219 :     Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<TL_SHIFT) + dDist21;
1220 :    
1221 :     if (Cost2<Cost1) {
1222 :     Cost1 = Cost2;
1223 :     bLevel = Level2;
1224 :     } else
1225 :     bLevel = Level1;
1226 :    
1227 :     if (Cost1<Last_Cost) {
1228 :     Last_Cost = Cost1;
1229 :     Last.Run = Run;
1230 :     Last.Level = bLevel;
1231 :     Last_Node = i;
1232 :     }
1233 :     } /* end of "for Run" */
1234 :    
1235 :     if (DBG==1) {
1236 :     Run_Costs[i] = Best_Cost;
1237 :     printf( "Costs #%2d: ", i);
1238 :     for(j=-1;j<=Non_Zero;++j) {
1239 :     if (j==Run_Start) printf( " %3.0d|", Run_Costs[j]>>12 );
1240 :     else if (j>Run_Start && j<i) printf( " %3.0d|", Run_Costs[j]>>12 );
1241 :     else if (j==i) printf( "(%3.0d)", Run_Costs[j]>>12 );
1242 :     else printf( " - |" );
1243 :     }
1244 :     printf( "<%3.0d %2d %d>", Min_Cost>>12, Nodes[i].Level, Nodes[i].Run );
1245 :     printf( " Last:#%2d {%3.0d %2d %d}", Last_Node, Last_Cost>>12, Last.Level, Last.Run );
1246 :     printf( " AC:%3.0d Dist0:%3d Dist(%2d):%3d Dist(%2d):%3d", AC, Dist0>>12, Level1, Dist1>>12, Level2, Dist2>>12 );
1247 :     printf( "\n" );
1248 :     }
1249 :     }
1250 :    
1251 :     Run_Costs[i] = Best_Cost;
1252 :    
1253 :     if (Best_Cost < Min_Cost + Dist0) {
1254 :     Min_Cost = Best_Cost;
1255 :     Run_Start = i;
1256 :     }
1257 :     else
1258 :     {
1259 :     /*
1260 :     * as noticed by Michael Niedermayer (michaelni at gmx.at), there's
1261 :     * a code shorter by 1 bit for a larger run (!), same level. We give
1262 :     * it a chance by not moving the left barrier too much.
1263 :     */
1264 :    
1265 :     while( Run_Costs[Run_Start]>Min_Cost+(1<<TL_SHIFT) )
1266 :     Run_Start++;
1267 :    
1268 :     /* spread on preceding coeffs the cost incurred by skipping this one */
1269 :     for(j=Run_Start; j<i; ++j) Run_Costs[j] += Dist0;
1270 :     Min_Cost += Dist0;
1271 :     }
1272 :     }
1273 :    
1274 :     if (DBG) {
1275 :     Last_Cost = Evaluate_Cost(Out,Mult,Bias, Zigzag,Non_Zero, Lambda);
1276 :     if (DBG==1) {
1277 :     printf( "=> " );
1278 :     for(i=0; i<=Non_Zero; ++i) printf( "[%3.0d] ", Out[Zigzag[i]] );
1279 :     printf( "\n" );
1280 :     }
1281 :     }
1282 :    
1283 :     if (Last_Node<0)
1284 :     return -1;
1285 :    
1286 :     /* reconstruct optimal sequence backward with surviving paths */
1287 :     memset(Out, 0x00, 64*sizeof(*Out));
1288 :     Out[Zigzag[Last_Node]] = Last.Level;
1289 :     i = Last_Node - Last.Run;
1290 :     while(i>=0) {
1291 :     Out[Zigzag[i]] = Nodes[i].Level;
1292 :     i -= Nodes[i].Run;
1293 :     }
1294 :    
1295 :     if (DBG) {
1296 :     uint32_t Cost = Evaluate_Cost(Out,Mult,Bias, Zigzag,Non_Zero, Lambda);
1297 :     if (DBG==1) {
1298 :     printf( "<= " );
1299 :     for(i=0; i<=Last_Node; ++i) printf( "[%3.0d] ", Out[Zigzag[i]] );
1300 :     printf( "\n--------------------------------\n" );
1301 :     }
1302 :     if (Cost>Last_Cost) printf( "!!! %u > %u\n", Cost, Last_Cost );
1303 :     }
1304 :     return Last_Node;
1305 :     }
1306 :    
1307 :     #undef DBG
1308 :    
1309 :     #endif
No admin address has been configured
 ViewVC Help
Powered by ViewVC 1.0.4