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

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

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1432 - (view) (download)

1 : edgomez 1382 /*****************************************************************************
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 : edgomez 1432 * $Id: mbtransquant.c,v 1.24 2004-04-15 19:44:06 edgomez Exp $
25 : edgomez 1382 *
26 :     ****************************************************************************/
27 : Isibaar 3
28 : edgomez 1382 #include <stdio.h>
29 :     #include <stdlib.h>
30 : edgomez 78 #include <string.h>
31 :    
32 : Isibaar 3 #include "../portab.h"
33 :     #include "mbfunctions.h"
34 :    
35 :     #include "../global.h"
36 :     #include "mem_transfer.h"
37 :     #include "timer.h"
38 : edgomez 1382 #include "../bitstream/mbcoding.h"
39 :     #include "../bitstream/zigzag.h"
40 : Isibaar 3 #include "../dct/fdct.h"
41 :     #include "../dct/idct.h"
42 : edgomez 1382 #include "../quant/quant.h"
43 : Isibaar 3 #include "../encoder.h"
44 :    
45 : edgomez 851 #include "../image/reduced.h"
46 : edgomez 1382 #include "../quant/quant_matrix.h"
47 : Isibaar 3
48 : edgomez 851 MBFIELDTEST_PTR MBFieldTest;
49 : Isibaar 3
50 : edgomez 1382 /*
51 :     * Skip blocks having a coefficient sum below this value. This value will be
52 :     * corrected according to the MB quantizer to avoid artifacts for quant==1
53 :     */
54 :     #define PVOP_TOOSMALL_LIMIT 1
55 :     #define BVOP_TOOSMALL_LIMIT 3
56 : Isibaar 3
57 : edgomez 1382 /*****************************************************************************
58 :     * Local functions
59 :     ****************************************************************************/
60 :    
61 :     /* permute block and return field dct choice */
62 :     static __inline uint32_t
63 :     MBDecideFieldDCT(int16_t data[6 * 64])
64 :     {
65 :     uint32_t field = MBFieldTest(data);
66 :    
67 :     if (field)
68 :     MBFrameToField(data);
69 :    
70 :     return field;
71 :     }
72 :    
73 :     /* Performs Forward DCT on all blocks */
74 : syskin 937 static __inline void
75 : edgomez 1382 MBfDCT(const MBParam * const pParam,
76 :     const FRAMEINFO * const frame,
77 :     MACROBLOCK * const pMB,
78 :     uint32_t x_pos,
79 :     uint32_t y_pos,
80 :     int16_t data[6 * 64])
81 : syskin 937 {
82 : edgomez 1382 /* Handles interlacing */
83 : syskin 937 start_timer();
84 : edgomez 1382 pMB->field_dct = 0;
85 :     if ((frame->vol_flags & XVID_VOL_INTERLACING) &&
86 :     (x_pos>0) && (x_pos<pParam->mb_width-1) &&
87 :     (y_pos>0) && (y_pos<pParam->mb_height-1)) {
88 :     pMB->field_dct = MBDecideFieldDCT(data);
89 :     }
90 :     stop_interlacing_timer();
91 :    
92 :     /* Perform DCT */
93 :     start_timer();
94 : syskin 937 fdct(&data[0 * 64]);
95 :     fdct(&data[1 * 64]);
96 :     fdct(&data[2 * 64]);
97 :     fdct(&data[3 * 64]);
98 :     fdct(&data[4 * 64]);
99 :     fdct(&data[5 * 64]);
100 :     stop_dct_timer();
101 :     }
102 :    
103 : edgomez 1382 /* Performs Inverse DCT on all blocks */
104 :     static __inline void
105 :     MBiDCT(int16_t data[6 * 64],
106 :     const uint8_t cbp)
107 :     {
108 :     start_timer();
109 :     if(cbp & (1 << (5 - 0))) idct(&data[0 * 64]);
110 :     if(cbp & (1 << (5 - 1))) idct(&data[1 * 64]);
111 :     if(cbp & (1 << (5 - 2))) idct(&data[2 * 64]);
112 :     if(cbp & (1 << (5 - 3))) idct(&data[3 * 64]);
113 :     if(cbp & (1 << (5 - 4))) idct(&data[4 * 64]);
114 :     if(cbp & (1 << (5 - 5))) idct(&data[5 * 64]);
115 :     stop_idct_timer();
116 :     }
117 : syskin 937
118 : edgomez 1382 /* Quantize all blocks -- Intra mode */
119 :     static __inline void
120 :     MBQuantIntra(const MBParam * pParam,
121 :     const FRAMEINFO * const frame,
122 :     const MACROBLOCK * pMB,
123 :     int16_t qcoeff[6 * 64],
124 :     int16_t data[6*64])
125 : syskin 937 {
126 : edgomez 1382 int mpeg;
127 :     int scaler_lum, scaler_chr;
128 : syskin 937
129 : edgomez 1382 quant_intraFuncPtr const quant[2] =
130 :     {
131 :     quant_h263_intra,
132 :     quant_mpeg_intra
133 :     };
134 :    
135 :     mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
136 :     scaler_lum = get_dc_scaler(pMB->quant, 1);
137 :     scaler_chr = get_dc_scaler(pMB->quant, 0);
138 :    
139 :     /* Quantize the block */
140 : syskin 937 start_timer();
141 : edgomez 1382 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, pParam->mpeg_quant_matrices);
143 :     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, pParam->mpeg_quant_matrices);
145 :     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, pParam->mpeg_quant_matrices);
147 : syskin 937 stop_quant_timer();
148 :     }
149 :    
150 : edgomez 1382 /* DeQuantize all blocks -- Intra mode */
151 :     static __inline void
152 :     MBDeQuantIntra(const MBParam * pParam,
153 :     const int iQuant,
154 :     int16_t qcoeff[6 * 64],
155 :     int16_t data[6*64])
156 : Isibaar 3 {
157 : edgomez 1382 int mpeg;
158 :     int scaler_lum, scaler_chr;
159 : edgomez 78
160 : edgomez 1382 quant_intraFuncPtr const dequant[2] =
161 :     {
162 :     dequant_h263_intra,
163 :     dequant_mpeg_intra
164 :     };
165 : Isibaar 3
166 : edgomez 1382 mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
167 :     scaler_lum = get_dc_scaler(iQuant, 1);
168 :     scaler_chr = get_dc_scaler(iQuant, 0);
169 :    
170 : edgomez 851 start_timer();
171 : edgomez 1382 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, pParam->mpeg_quant_matrices);
173 :     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, pParam->mpeg_quant_matrices);
175 :     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, pParam->mpeg_quant_matrices);
177 :     stop_iquant_timer();
178 :     }
179 : h 69
180 : edgomez 1382 static int
181 :     dct_quantize_trellis_c(int16_t *const Out,
182 :     const int16_t *const In,
183 :     int Q,
184 :     const uint16_t * const Zigzag,
185 :     const uint16_t * const QuantMatrix,
186 : edgomez 1432 int Non_Zero,
187 :     int Sum);
188 : edgomez 851
189 : edgomez 1382 /* 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 : h 69
200 : edgomez 1382 int i;
201 :     uint8_t cbp = 0;
202 :     int sum;
203 :     int code_block, mpeg;
204 : h 69
205 : edgomez 1382 quant_interFuncPtr const quant[2] =
206 :     {
207 :     quant_h263_inter,
208 :     quant_mpeg_inter
209 :     };
210 : syskin 937
211 : edgomez 1382 mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
212 :    
213 : edgomez 195 for (i = 0; i < 6; i++) {
214 : Isibaar 3
215 : edgomez 1382 /* Quantize the block */
216 : Isibaar 3 start_timer();
217 :    
218 : edgomez 1382 sum = quant[mpeg](&qcoeff[i*64], &data[i*64], pMB->quant, pParam->mpeg_quant_matrices);
219 : edgomez 851
220 : edgomez 1382 if(sum && (frame->vop_flags & XVID_VOP_TRELLISQUANT)) {
221 :     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 1432 63,
239 :     sum);
240 : Isibaar 3 }
241 : edgomez 1382 stop_quant_timer();
242 : Isibaar 3
243 : edgomez 1382 /*
244 :     * We code the block if the sum is higher than the limit and if the first
245 :     * two AC coefficients in zig zag order are not zero.
246 :     */
247 :     code_block = 0;
248 :     if ((sum >= limit) || (qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) {
249 :     code_block = 1;
250 :     } else {
251 : edgomez 851
252 : edgomez 1382 if (bvop && (pMB->mode == MODE_DIRECT || pMB->mode == MODE_DIRECT_NO4V)) {
253 :     /* dark blocks prevention for direct mode */
254 :     if ((qcoeff[i*64] < -1) || (qcoeff[i*64] > 0))
255 :     code_block = 1;
256 :     } else {
257 :     /* not direct mode */
258 :     if (qcoeff[i*64] != 0)
259 :     code_block = 1;
260 :     }
261 : edgomez 851 }
262 :    
263 : edgomez 1382 /* Set the corresponding cbp bit */
264 :     cbp |= code_block << (5 - i);
265 : h 69 }
266 :    
267 : edgomez 1382 return(cbp);
268 : Isibaar 3 }
269 :    
270 : edgomez 1382 /* DeQuantize all blocks -- Inter mode */
271 :     static __inline void
272 :     MBDeQuantInter(const MBParam * pParam,
273 :     const int iQuant,
274 :     int16_t data[6 * 64],
275 :     int16_t qcoeff[6 * 64],
276 :     const uint8_t cbp)
277 : Isibaar 3 {
278 : edgomez 1382 int mpeg;
279 :    
280 :     quant_interFuncPtr const dequant[2] =
281 :     {
282 :     dequant_h263_inter,
283 :     dequant_mpeg_inter
284 :     };
285 :    
286 :     mpeg = !!(pParam->vol_flags & XVID_VOL_MPEGQUANT);
287 :    
288 :     start_timer();
289 :     if(cbp & (1 << (5 - 0))) dequant[mpeg](&data[0 * 64], &qcoeff[0 * 64], iQuant, pParam->mpeg_quant_matrices);
290 :     if(cbp & (1 << (5 - 1))) dequant[mpeg](&data[1 * 64], &qcoeff[1 * 64], iQuant, pParam->mpeg_quant_matrices);
291 :     if(cbp & (1 << (5 - 2))) dequant[mpeg](&data[2 * 64], &qcoeff[2 * 64], iQuant, pParam->mpeg_quant_matrices);
292 :     if(cbp & (1 << (5 - 3))) dequant[mpeg](&data[3 * 64], &qcoeff[3 * 64], iQuant, pParam->mpeg_quant_matrices);
293 :     if(cbp & (1 << (5 - 4))) dequant[mpeg](&data[4 * 64], &qcoeff[4 * 64], iQuant, pParam->mpeg_quant_matrices);
294 :     if(cbp & (1 << (5 - 5))) dequant[mpeg](&data[5 * 64], &qcoeff[5 * 64], iQuant, pParam->mpeg_quant_matrices);
295 :     stop_iquant_timer();
296 :     }
297 :    
298 :     typedef void (transfer_operation_8to16_t) (int16_t *Dst, const uint8_t *Src, int BpS);
299 :     typedef void (transfer_operation_16to8_t) (uint8_t *Dst, const int16_t *Src, int BpS);
300 :    
301 :    
302 :     static __inline void
303 :     MBTrans8to16(const MBParam * const pParam,
304 :     const FRAMEINFO * const frame,
305 :     const MACROBLOCK * const pMB,
306 :     const uint32_t x_pos,
307 :     const uint32_t y_pos,
308 :     int16_t data[6 * 64])
309 :     {
310 : h 82 uint32_t stride = pParam->edged_width;
311 : edgomez 1382 uint32_t stride2 = stride / 2;
312 :     uint32_t next_block = stride * 8;
313 :     int32_t cst;
314 :     int vop_reduced;
315 : Isibaar 3 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
316 : syskin 937 const IMAGE * const pCurrent = &frame->image;
317 : edgomez 1382 transfer_operation_8to16_t * const functions[2] =
318 :     {
319 :     (transfer_operation_8to16_t *)transfer_8to16copy,
320 :     (transfer_operation_8to16_t *)filter_18x18_to_8x8
321 :     };
322 :     transfer_operation_8to16_t *transfer_op = NULL;
323 : edgomez 195
324 : edgomez 1382 vop_reduced = !!(frame->vop_flags & XVID_VOP_REDUCED);
325 : Isibaar 3
326 : edgomez 1382 /* Image pointers */
327 :     pY_Cur = pCurrent->y + (y_pos << (4+vop_reduced)) * stride + (x_pos << (4+vop_reduced));
328 :     pU_Cur = pCurrent->u + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced));
329 :     pV_Cur = pCurrent->v + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced));
330 : h 69
331 : edgomez 1382 /* Block size */
332 :     cst = 8<<vop_reduced;
333 : edgomez 851
334 : edgomez 1382 /* Operation function */
335 :     transfer_op = functions[vop_reduced];
336 : Isibaar 3
337 : edgomez 1382 /* Do the transfer */
338 :     start_timer();
339 :     transfer_op(&data[0 * 64], pY_Cur, stride);
340 :     transfer_op(&data[1 * 64], pY_Cur + cst, stride);
341 :     transfer_op(&data[2 * 64], pY_Cur + next_block, stride);
342 :     transfer_op(&data[3 * 64], pY_Cur + next_block + cst, stride);
343 :     transfer_op(&data[4 * 64], pU_Cur, stride2);
344 :     transfer_op(&data[5 * 64], pV_Cur, stride2);
345 :     stop_transfer_timer();
346 :     }
347 : Isibaar 3
348 : edgomez 1382 static __inline void
349 :     MBTrans16to8(const MBParam * const pParam,
350 :     const FRAMEINFO * const frame,
351 :     const MACROBLOCK * const pMB,
352 :     const uint32_t x_pos,
353 :     const uint32_t y_pos,
354 :     int16_t data[6 * 64],
355 :     const uint32_t add, /* Must be 1 or 0 */
356 :     const uint8_t cbp)
357 :     {
358 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
359 :     uint32_t stride = pParam->edged_width;
360 :     uint32_t stride2 = stride / 2;
361 :     uint32_t next_block = stride * 8;
362 :     uint32_t cst;
363 :     int vop_reduced;
364 :     const IMAGE * const pCurrent = &frame->image;
365 : Isibaar 3
366 : edgomez 1382 /* Array of function pointers, indexed by [vop_reduced<<1+add] */
367 :     transfer_operation_16to8_t * const functions[4] =
368 :     {
369 :     (transfer_operation_16to8_t*)transfer_16to8copy,
370 :     (transfer_operation_16to8_t*)transfer_16to8add,
371 :     (transfer_operation_16to8_t*)copy_upsampled_8x8_16to8,
372 :     (transfer_operation_16to8_t*)add_upsampled_8x8_16to8
373 :     };
374 : Isibaar 3
375 : edgomez 1382 transfer_operation_16to8_t *transfer_op = NULL;
376 : Isibaar 3
377 : edgomez 1382 /* Makes this vars booleans */
378 :     vop_reduced = !!(frame->vop_flags & XVID_VOP_REDUCED);
379 : h 69
380 : edgomez 1382 /* Image pointers */
381 :     pY_Cur = pCurrent->y + (y_pos << (4+vop_reduced)) * stride + (x_pos << (4+vop_reduced));
382 :     pU_Cur = pCurrent->u + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced));
383 :     pV_Cur = pCurrent->v + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced));
384 :    
385 : edgomez 195 if (pMB->field_dct) {
386 : h 82 next_block = stride;
387 :     stride *= 2;
388 : h 69 }
389 :    
390 : edgomez 1382 /* Block size */
391 :     cst = 8<<vop_reduced;
392 :    
393 :     /* Operation function */
394 :     transfer_op = functions[(vop_reduced<<1) + add];
395 :    
396 :     /* Do the operation */
397 : h 69 start_timer();
398 : edgomez 1382 if (cbp&32) transfer_op(pY_Cur, &data[0 * 64], stride);
399 :     if (cbp&16) transfer_op(pY_Cur + cst, &data[1 * 64], stride);
400 :     if (cbp& 8) transfer_op(pY_Cur + next_block, &data[2 * 64], stride);
401 :     if (cbp& 4) transfer_op(pY_Cur + next_block + cst, &data[3 * 64], stride);
402 :     if (cbp& 2) transfer_op(pU_Cur, &data[4 * 64], stride2);
403 :     if (cbp& 1) transfer_op(pV_Cur, &data[5 * 64], stride2);
404 : h 69 stop_transfer_timer();
405 : edgomez 1382 }
406 : h 69
407 : edgomez 1382 /*****************************************************************************
408 :     * Module functions
409 :     ****************************************************************************/
410 :    
411 :     void
412 :     MBTransQuantIntra(const MBParam * const pParam,
413 :     const FRAMEINFO * const frame,
414 :     MACROBLOCK * const pMB,
415 :     const uint32_t x_pos,
416 :     const uint32_t y_pos,
417 :     int16_t data[6 * 64],
418 :     int16_t qcoeff[6 * 64])
419 :     {
420 :    
421 :     /* Transfer data */
422 :     MBTrans8to16(pParam, frame, pMB, x_pos, y_pos, data);
423 :    
424 :     /* Perform DCT (and field decision) */
425 :     MBfDCT(pParam, frame, pMB, x_pos, y_pos, data);
426 :    
427 :     /* Quantize the block */
428 :     MBQuantIntra(pParam, frame, pMB, data, qcoeff);
429 :    
430 :     /* DeQuantize the block */
431 :     MBDeQuantIntra(pParam, pMB->quant, data, qcoeff);
432 :    
433 :     /* Perform inverse DCT*/
434 :     MBiDCT(data, 0x3F);
435 :    
436 :     /* Transfer back the data -- Don't add data */
437 :     MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 0, 0x3F);
438 : Isibaar 3 }
439 : h 69
440 : edgomez 1382
441 : chl 368 uint8_t
442 : edgomez 1382 MBTransQuantInter(const MBParam * const pParam,
443 :     const FRAMEINFO * const frame,
444 :     MACROBLOCK * const pMB,
445 :     const uint32_t x_pos,
446 :     const uint32_t y_pos,
447 : chl 368 int16_t data[6 * 64],
448 :     int16_t qcoeff[6 * 64])
449 :     {
450 : edgomez 1382 uint8_t cbp;
451 :     uint32_t limit;
452 : chl 368
453 : edgomez 1382 /* There is no MBTrans8to16 for Inter block, that's done in motion compensation
454 :     * already */
455 : chl 368
456 : edgomez 1382 /* Perform DCT (and field decision) */
457 :     MBfDCT(pParam, frame, pMB, x_pos, y_pos, data);
458 : chl 368
459 : edgomez 1382 /* Set the limit threshold */
460 :     limit = PVOP_TOOSMALL_LIMIT + ((pMB->quant == 1)? 1 : 0);
461 : chl 368
462 : edgomez 1382 if (frame->vop_flags & XVID_VOP_CARTOON)
463 :     limit *= 3;
464 : chl 368
465 : edgomez 1382 /* Quantize the block */
466 :     cbp = MBQuantInter(pParam, frame, pMB, data, qcoeff, 0, limit);
467 : chl 368
468 : edgomez 1382 /* DeQuantize the block */
469 :     MBDeQuantInter(pParam, pMB->quant, data, qcoeff, cbp);
470 :    
471 :     /* Perform inverse DCT*/
472 :     MBiDCT(data, cbp);
473 :    
474 :     /* Transfer back the data -- Add the data */
475 :     MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 1, cbp);
476 :    
477 :     return(cbp);
478 : chl 368 }
479 :    
480 : edgomez 1382 uint8_t
481 :     MBTransQuantInterBVOP(const MBParam * pParam,
482 :     FRAMEINFO * frame,
483 :     MACROBLOCK * pMB,
484 :     const uint32_t x_pos,
485 :     const uint32_t y_pos,
486 :     int16_t data[6 * 64],
487 :     int16_t qcoeff[6 * 64])
488 : h 69 {
489 : edgomez 1382 uint8_t cbp;
490 :     uint32_t limit;
491 : h 69
492 : edgomez 1382 /* There is no MBTrans8to16 for Inter block, that's done in motion compensation
493 :     * already */
494 :    
495 :     /* Perform DCT (and field decision) */
496 :     MBfDCT(pParam, frame, pMB, x_pos, y_pos, data);
497 :    
498 :     /* Set the limit threshold */
499 :     limit = BVOP_TOOSMALL_LIMIT;
500 :    
501 :     if (frame->vop_flags & XVID_VOP_CARTOON)
502 :     limit *= 2;
503 :    
504 :     /* Quantize the block */
505 :     cbp = MBQuantInter(pParam, frame, pMB, data, qcoeff, 1, limit);
506 :    
507 :     /*
508 :     * History comment:
509 :     * We don't have to DeQuant, iDCT and Transfer back data for B-frames.
510 :     *
511 :     * BUT some plugins require the rebuilt original frame to be passed so we
512 :     * have to take care of that here
513 :     */
514 :     if((pParam->plugin_flags & XVID_REQORIGINAL)) {
515 :    
516 :     /* DeQuantize the block */
517 :     MBDeQuantInter(pParam, pMB->quant, data, qcoeff, cbp);
518 :    
519 :     /* Perform inverse DCT*/
520 :     MBiDCT(data, cbp);
521 :    
522 :     /* Transfer back the data -- Add the data */
523 :     MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 1, cbp);
524 :     }
525 :    
526 :     return(cbp);
527 : edgomez 851 }
528 :    
529 :     /* if sum(diff between field lines) < sum(diff between frame lines), use field dct */
530 :     uint32_t
531 :     MBFieldTest_c(int16_t data[6 * 64])
532 :     {
533 : edgomez 195 const uint8_t blocks[] =
534 :     { 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 };
535 :     const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 };
536 : edgomez 78
537 : h 69 int frame = 0, field = 0;
538 :     int i, j;
539 :    
540 : edgomez 195 for (i = 0; i < 7; ++i) {
541 :     for (j = 0; j < 8; ++j) {
542 :     frame +=
543 : edgomez 1382 abs(data[0 * 64 + (i + 1) * 8 + j] - data[0 * 64 + i * 8 + j]);
544 : edgomez 195 frame +=
545 : edgomez 1382 abs(data[1 * 64 + (i + 1) * 8 + j] - data[1 * 64 + i * 8 + j]);
546 : edgomez 195 frame +=
547 : edgomez 1382 abs(data[2 * 64 + (i + 1) * 8 + j] - data[2 * 64 + i * 8 + j]);
548 : edgomez 195 frame +=
549 : edgomez 1382 abs(data[3 * 64 + (i + 1) * 8 + j] - data[3 * 64 + i * 8 + j]);
550 : h 69
551 : edgomez 195 field +=
552 : edgomez 1382 abs(data[blocks[i + 1] + lines[i + 1] + j] -
553 : edgomez 195 data[blocks[i] + lines[i] + j]);
554 :     field +=
555 : edgomez 1382 abs(data[blocks[i + 1] + lines[i + 1] + 8 + j] -
556 : edgomez 195 data[blocks[i] + lines[i] + 8 + j]);
557 :     field +=
558 : edgomez 1382 abs(data[blocks[i + 1] + 64 + lines[i + 1] + j] -
559 : edgomez 195 data[blocks[i] + 64 + lines[i] + j]);
560 :     field +=
561 : edgomez 1382 abs(data[blocks[i + 1] + 64 + lines[i + 1] + 8 + j] -
562 : edgomez 195 data[blocks[i] + 64 + lines[i] + 8 + j]);
563 : h 69 }
564 :     }
565 :    
566 : edgomez 851 return (frame >= (field + 350));
567 : h 69 }
568 :    
569 :    
570 :     /* deinterlace Y blocks vertically */
571 :    
572 :     #define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp))
573 : edgomez 1382 #define LINE(X,Y) &data[X*64 + Y*8]
574 : h 69
575 : edgomez 195 void
576 :     MBFrameToField(int16_t data[6 * 64])
577 : h 69 {
578 :     int16_t tmp[8];
579 :    
580 :     /* left blocks */
581 :    
582 : edgomez 1382 /* 1=2, 2=4, 4=8, 8=1 */
583 : edgomez 195 MOVLINE(tmp, LINE(0, 1));
584 :     MOVLINE(LINE(0, 1), LINE(0, 2));
585 :     MOVLINE(LINE(0, 2), LINE(0, 4));
586 :     MOVLINE(LINE(0, 4), LINE(2, 0));
587 :     MOVLINE(LINE(2, 0), tmp);
588 : h 69
589 : edgomez 1382 /* 3=6, 6=12, 12=9, 9=3 */
590 : edgomez 195 MOVLINE(tmp, LINE(0, 3));
591 :     MOVLINE(LINE(0, 3), LINE(0, 6));
592 :     MOVLINE(LINE(0, 6), LINE(2, 4));
593 :     MOVLINE(LINE(2, 4), LINE(2, 1));
594 :     MOVLINE(LINE(2, 1), tmp);
595 : h 69
596 : edgomez 1382 /* 5=10, 10=5 */
597 : edgomez 195 MOVLINE(tmp, LINE(0, 5));
598 :     MOVLINE(LINE(0, 5), LINE(2, 2));
599 :     MOVLINE(LINE(2, 2), tmp);
600 : h 69
601 : edgomez 1382 /* 7=14, 14=13, 13=11, 11=7 */
602 : edgomez 195 MOVLINE(tmp, LINE(0, 7));
603 :     MOVLINE(LINE(0, 7), LINE(2, 6));
604 :     MOVLINE(LINE(2, 6), LINE(2, 5));
605 :     MOVLINE(LINE(2, 5), LINE(2, 3));
606 :     MOVLINE(LINE(2, 3), tmp);
607 : h 69
608 :     /* right blocks */
609 :    
610 : edgomez 1382 /* 1=2, 2=4, 4=8, 8=1 */
611 : edgomez 195 MOVLINE(tmp, LINE(1, 1));
612 :     MOVLINE(LINE(1, 1), LINE(1, 2));
613 :     MOVLINE(LINE(1, 2), LINE(1, 4));
614 :     MOVLINE(LINE(1, 4), LINE(3, 0));
615 :     MOVLINE(LINE(3, 0), tmp);
616 : h 69
617 : edgomez 1382 /* 3=6, 6=12, 12=9, 9=3 */
618 : edgomez 195 MOVLINE(tmp, LINE(1, 3));
619 :     MOVLINE(LINE(1, 3), LINE(1, 6));
620 :     MOVLINE(LINE(1, 6), LINE(3, 4));
621 :     MOVLINE(LINE(3, 4), LINE(3, 1));
622 :     MOVLINE(LINE(3, 1), tmp);
623 : h 69
624 : edgomez 1382 /* 5=10, 10=5 */
625 : edgomez 195 MOVLINE(tmp, LINE(1, 5));
626 :     MOVLINE(LINE(1, 5), LINE(3, 2));
627 :     MOVLINE(LINE(3, 2), tmp);
628 : h 69
629 : edgomez 1382 /* 7=14, 14=13, 13=11, 11=7 */
630 : edgomez 195 MOVLINE(tmp, LINE(1, 7));
631 :     MOVLINE(LINE(1, 7), LINE(3, 6));
632 :     MOVLINE(LINE(3, 6), LINE(3, 5));
633 :     MOVLINE(LINE(3, 5), LINE(3, 3));
634 :     MOVLINE(LINE(3, 3), tmp);
635 : h 69 }
636 : edgomez 1382
637 :     /*****************************************************************************
638 :     * Trellis based R-D optimal quantization
639 :     *
640 :     * Trellis Quant code (C) 2003 Pascal Massimino skal(at)planet-d.net
641 :     *
642 :     ****************************************************************************/
643 :    
644 :     /*----------------------------------------------------------------------------
645 :     *
646 :     * Trellis-Based quantization
647 :     *
648 :     * So far I understand this paper:
649 :     *
650 :     * "Trellis-Based R-D Optimal Quantization in H.263+"
651 :     * J.Wen, M.Luttrell, J.Villasenor
652 :     * IEEE Transactions on Image Processing, Vol.9, No.8, Aug. 2000.
653 :     *
654 :     * we are at stake with a simplified Bellmand-Ford / Dijkstra Single
655 :     * Source Shortest Path algo. But due to the underlying graph structure
656 :     * ("Trellis"), it can be turned into a dynamic programming algo,
657 :     * partially saving the explicit graph's nodes representation. And
658 :     * without using a heap, since the open frontier of the DAG is always
659 :     * known, and of fixed size.
660 :     *--------------------------------------------------------------------------*/
661 :    
662 :    
663 :    
664 :     /* Codes lengths for relevant levels. */
665 :    
666 :     /* let's factorize: */
667 :     static const uint8_t Code_Len0[64] = {
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 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
670 :     static const uint8_t Code_Len1[64] = {
671 :     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,
672 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
673 :     static const uint8_t Code_Len2[64] = {
674 :     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,
675 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
676 :     static const uint8_t Code_Len3[64] = {
677 :     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,
678 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
679 :     static const uint8_t Code_Len4[64] = {
680 :     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,
681 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
682 :     static const uint8_t Code_Len5[64] = {
683 :     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,
684 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
685 :     static const uint8_t Code_Len6[64] = {
686 :     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,
687 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
688 :     static const uint8_t Code_Len7[64] = {
689 :     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,
690 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
691 :     static const uint8_t Code_Len8[64] = {
692 :     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,
693 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
694 :     static const uint8_t Code_Len9[64] = {
695 :     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,
696 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
697 :     static const uint8_t Code_Len10[64] = {
698 :     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,
699 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
700 :     static const uint8_t Code_Len11[64] = {
701 :     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,
702 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
703 :     static const uint8_t Code_Len12[64] = {
704 :     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,
705 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
706 :     static const uint8_t Code_Len13[64] = {
707 :     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,
708 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
709 :     static const uint8_t Code_Len14[64] = {
710 :     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,
711 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
712 :     static const uint8_t Code_Len15[64] = {
713 :     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,
714 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
715 :     static const uint8_t Code_Len16[64] = {
716 :     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,
717 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30};
718 :     static const uint8_t Code_Len17[64] = {
719 :     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,
720 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
721 :     static const uint8_t Code_Len18[64] = {
722 :     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,
723 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
724 :     static const uint8_t Code_Len19[64] = {
725 :     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,
726 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30 };
727 :     static const uint8_t Code_Len20[64] = {
728 :     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,
729 :     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 };
730 :    
731 :     /* a few more table for LAST table: */
732 :     static const uint8_t Code_Len21[64] = {
733 :     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,
734 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30};
735 :     static const uint8_t Code_Len22[64] = {
736 :     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,
737 :     30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30,30};
738 :     static const uint8_t Code_Len23[64] = {
739 :     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,
740 :     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};
741 :     static const uint8_t Code_Len24[64] = {
742 :     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,
743 :     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};
744 :    
745 :    
746 :     static const uint8_t * const B16_17_Code_Len[24] = { /* levels [1..24] */
747 :     Code_Len20,Code_Len19,Code_Len18,Code_Len17,
748 :     Code_Len16,Code_Len15,Code_Len14,Code_Len13,
749 :     Code_Len12,Code_Len11,Code_Len10,Code_Len9,
750 :     Code_Len8, Code_Len7 ,Code_Len6 ,Code_Len5,
751 :     Code_Len4, Code_Len3, Code_Len3 ,Code_Len2,
752 :     Code_Len2, Code_Len1, Code_Len1, Code_Len1,
753 :     };
754 :    
755 :     static const uint8_t * const B16_17_Code_Len_Last[6] = { /* levels [1..6] */
756 :     Code_Len24,Code_Len23,Code_Len22,Code_Len21, Code_Len3, Code_Len1,
757 :     };
758 :    
759 :     /* TL_SHIFT controls the precision of the RD optimizations in trellis
760 :     * valid range is [10..16]. The bigger, the more trellis is vulnerable
761 :     * to overflows in cost formulas.
762 :     * - 10 allows ac values up to 2^11 == 2048
763 :     * - 16 allows ac values up to 2^8 == 256
764 :     */
765 :     #define TL_SHIFT 11
766 :     #define TL(q) ((0xfe00>>(16-TL_SHIFT))/(q*q))
767 :    
768 :     static const int Trellis_Lambda_Tabs[31] = {
769 :     TL( 1),TL( 2),TL( 3),TL( 4),TL( 5),TL( 6), TL( 7),
770 :     TL( 8),TL( 9),TL(10),TL(11),TL(12),TL(13),TL(14), TL(15),
771 :     TL(16),TL(17),TL(18),TL(19),TL(20),TL(21),TL(22), TL(23),
772 :     TL(24),TL(25),TL(26),TL(27),TL(28),TL(29),TL(30), TL(31)
773 :     };
774 :     #undef TL
775 :    
776 :     static int __inline
777 :     Find_Last(const int16_t *C, const uint16_t *Zigzag, int i)
778 :     {
779 :     while(i>=0)
780 :     if (C[Zigzag[i]])
781 :     return i;
782 :     else i--;
783 :     return -1;
784 :     }
785 :    
786 :     /* this routine has been strippen of all debug code */
787 :     static int
788 :     dct_quantize_trellis_c(int16_t *const Out,
789 :     const int16_t *const In,
790 :     int Q,
791 :     const uint16_t * const Zigzag,
792 :     const uint16_t * const QuantMatrix,
793 : edgomez 1432 int Non_Zero,
794 :     int Sum)
795 : edgomez 1382 {
796 :    
797 :     /* Note: We should search last non-zero coeffs on *real* DCT input coeffs
798 :     * (In[]), not quantized one (Out[]). However, it only improves the result
799 :     * *very* slightly (~0.01dB), whereas speed drops to crawling level :)
800 :     * Well, actually, taking 1 more coeff past Non_Zero into account sometimes
801 :     * helps. */
802 :     typedef struct { int16_t Run, Level; } NODE;
803 :    
804 :     NODE Nodes[65], Last;
805 :     uint32_t Run_Costs0[64+1];
806 :     uint32_t * const Run_Costs = Run_Costs0 + 1;
807 :    
808 :     /* it's 1/lambda, actually */
809 :     const int Lambda = Trellis_Lambda_Tabs[Q-1];
810 :    
811 :     int Run_Start = -1;
812 :     uint32_t Min_Cost = 2<<TL_SHIFT;
813 :    
814 :     int Last_Node = -1;
815 :     uint32_t Last_Cost = 0;
816 :    
817 : edgomez 1432 int i, j;
818 : edgomez 1382
819 :     /* source (w/ CBP penalty) */
820 :     Run_Costs[-1] = 2<<TL_SHIFT;
821 :    
822 :     Non_Zero = Find_Last(Out, Zigzag, Non_Zero);
823 :     if (Non_Zero<0)
824 :     return 0; /* Sum is zero if there are only zero coeffs */
825 :    
826 :     for(i=0; i<=Non_Zero; i++) {
827 :     const int q = ((Q*QuantMatrix[Zigzag[i]])>>4);
828 :     const int Mult = 2*q;
829 :     const int Bias = (q-1) | 1;
830 :     const int Lev0 = Mult + Bias;
831 :    
832 :     const int AC = In[Zigzag[i]];
833 :     const int Level1 = Out[Zigzag[i]];
834 :     const unsigned int Dist0 = Lambda* AC*AC;
835 :     uint32_t Best_Cost = 0xf0000000;
836 :     Last_Cost += Dist0;
837 :    
838 :     /* very specialized loop for -1,0,+1 */
839 :     if ((uint32_t)(Level1+1)<3) {
840 :     int dQ;
841 :     int Run;
842 :     uint32_t Cost0;
843 :    
844 :     if (AC<0) {
845 :     Nodes[i].Level = -1;
846 :     dQ = Lev0 + AC;
847 :     } else {
848 :     Nodes[i].Level = 1;
849 :     dQ = Lev0 - AC;
850 :     }
851 :     Cost0 = Lambda*dQ*dQ;
852 :    
853 :     Nodes[i].Run = 1;
854 :     Best_Cost = (Code_Len20[0]<<TL_SHIFT) + Run_Costs[i-1]+Cost0;
855 :     for(Run=i-Run_Start; Run>0; --Run) {
856 :     const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run];
857 :     const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<TL_SHIFT);
858 :     const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<TL_SHIFT);
859 :    
860 :     /* TODO: what about tie-breaks? Should we favor short runs or
861 :     * long runs? Although the error is the same, it would not be
862 :     * spread the same way along high and low frequencies... */
863 :    
864 :     /* Gruel: I'd say, favour short runs => hifreq errors (HVS) */
865 :    
866 :     if (Cost<Best_Cost) {
867 :     Best_Cost = Cost;
868 :     Nodes[i].Run = Run;
869 :     }
870 :    
871 :     if (lCost<Last_Cost) {
872 :     Last_Cost = lCost;
873 :     Last.Run = Run;
874 :     Last_Node = i;
875 :     }
876 :     }
877 :     if (Last_Node==i)
878 :     Last.Level = Nodes[i].Level;
879 :     } else if (51U>(uint32_t)(Level1+25)) {
880 :     /* "big" levels (not less than ESC3, though) */
881 :     const uint8_t *Tbl_L1, *Tbl_L2, *Tbl_L1_Last, *Tbl_L2_Last;
882 :     int Level2;
883 :     int dQ1, dQ2;
884 :     int Run;
885 :     uint32_t Dist1,Dist2;
886 :     int dDist21;
887 :    
888 :     if (Level1>1) {
889 :     dQ1 = Level1*Mult-AC + Bias;
890 :     dQ2 = dQ1 - Mult;
891 :     Level2 = Level1-1;
892 :     Tbl_L1 = (Level1<=24) ? B16_17_Code_Len[Level1-1] : Code_Len0;
893 :     Tbl_L2 = (Level2<=24) ? B16_17_Code_Len[Level2-1] : Code_Len0;
894 :     Tbl_L1_Last = (Level1<=6) ? B16_17_Code_Len_Last[Level1-1] : Code_Len0;
895 :     Tbl_L2_Last = (Level2<=6) ? B16_17_Code_Len_Last[Level2-1] : Code_Len0;
896 :     } else { /* Level1<-1 */
897 :     dQ1 = Level1*Mult-AC - Bias;
898 :     dQ2 = dQ1 + Mult;
899 :     Level2 = Level1 + 1;
900 :     Tbl_L1 = (Level1>=-24) ? B16_17_Code_Len[Level1^-1] : Code_Len0;
901 :     Tbl_L2 = (Level2>=-24) ? B16_17_Code_Len[Level2^-1] : Code_Len0;
902 :     Tbl_L1_Last = (Level1>=- 6) ? B16_17_Code_Len_Last[Level1^-1] : Code_Len0;
903 :     Tbl_L2_Last = (Level2>=- 6) ? B16_17_Code_Len_Last[Level2^-1] : Code_Len0;
904 :     }
905 :    
906 :     Dist1 = Lambda*dQ1*dQ1;
907 :     Dist2 = Lambda*dQ2*dQ2;
908 :     dDist21 = Dist2-Dist1;
909 :    
910 :     for(Run=i-Run_Start; Run>0; --Run)
911 :     {
912 :     const uint32_t Cost_Base = Dist1 + Run_Costs[i-Run];
913 :     uint32_t Cost1, Cost2;
914 :     int bLevel;
915 :    
916 :     /* for sub-optimal (but slightly worth it, speed-wise) search,
917 :     * uncomment the following:
918 :     * if (Cost_Base>=Best_Cost) continue;
919 :     * (? doesn't seem to have any effect -- gruel ) */
920 :    
921 :     Cost1 = Cost_Base + (Tbl_L1[Run-1]<<TL_SHIFT);
922 :     Cost2 = Cost_Base + (Tbl_L2[Run-1]<<TL_SHIFT) + dDist21;
923 :    
924 :     if (Cost2<Cost1) {
925 :     Cost1 = Cost2;
926 :     bLevel = Level2;
927 :     } else {
928 :     bLevel = Level1;
929 :     }
930 :    
931 :     if (Cost1<Best_Cost) {
932 :     Best_Cost = Cost1;
933 :     Nodes[i].Run = Run;
934 :     Nodes[i].Level = bLevel;
935 :     }
936 :    
937 :     Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<TL_SHIFT);
938 :     Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<TL_SHIFT) + dDist21;
939 :    
940 :     if (Cost2<Cost1) {
941 :     Cost1 = Cost2;
942 :     bLevel = Level2;
943 :     } else {
944 :     bLevel = Level1;
945 :     }
946 :    
947 :     if (Cost1<Last_Cost) {
948 :     Last_Cost = Cost1;
949 :     Last.Run = Run;
950 :     Last.Level = bLevel;
951 :     Last_Node = i;
952 :     }
953 :     } /* end of "for Run" */
954 :     } else {
955 :     /* Very very high levels, with no chance of being optimizable
956 :     * => Simply pick best Run. */
957 :     int Run;
958 :     for(Run=i-Run_Start; Run>0; --Run) {
959 :     /* 30 bits + no distortion */
960 :     const uint32_t Cost = (30<<TL_SHIFT) + Run_Costs[i-Run];
961 :     if (Cost<Best_Cost) {
962 :     Best_Cost = Cost;
963 :     Nodes[i].Run = Run;
964 :     Nodes[i].Level = Level1;
965 :     }
966 :    
967 :     if (Cost<Last_Cost) {
968 :     Last_Cost = Cost;
969 :     Last.Run = Run;
970 :     Last.Level = Level1;
971 :     Last_Node = i;
972 :     }
973 :     }
974 :     }
975 :    
976 :    
977 :     Run_Costs[i] = Best_Cost;
978 :    
979 :     if (Best_Cost < Min_Cost + Dist0) {
980 :     Min_Cost = Best_Cost;
981 :     Run_Start = i;
982 :     } else {
983 :     /* as noticed by Michael Niedermayer (michaelni at gmx.at),
984 :     * there's a code shorter by 1 bit for a larger run (!), same
985 :     * level. We give it a chance by not moving the left barrier too
986 :     * much. */
987 :     while( Run_Costs[Run_Start]>Min_Cost+(1<<TL_SHIFT) )
988 :     Run_Start++;
989 :    
990 :     /* spread on preceding coeffs the cost incurred by skipping this
991 :     * one */
992 :     for(j=Run_Start; j<i; ++j) Run_Costs[j] += Dist0;
993 :     Min_Cost += Dist0;
994 :     }
995 :     }
996 :    
997 : edgomez 1432 /* It seems trellis doesn't give good results... just leave the block untouched
998 :     * and return the original sum value */
999 : edgomez 1382 if (Last_Node<0)
1000 : edgomez 1432 return Sum;
1001 : edgomez 1382
1002 :     /* reconstruct optimal sequence backward with surviving paths */
1003 :     memset(Out, 0x00, 64*sizeof(*Out));
1004 :     Out[Zigzag[Last_Node]] = Last.Level;
1005 :     i = Last_Node - Last.Run;
1006 : edgomez 1432 Sum = 0;
1007 : edgomez 1382 while(i>=0) {
1008 :     Out[Zigzag[i]] = Nodes[i].Level;
1009 : edgomez 1432 Sum += abs(Nodes[i].Level);
1010 : edgomez 1382 i -= Nodes[i].Run;
1011 :     }
1012 :    
1013 : edgomez 1432 return Sum;
1014 : edgomez 1382 }
1015 :    
1016 :     /* original version including heavy debugging info */
1017 :    
1018 :     #ifdef DBGTRELL
1019 :    
1020 :     #define DBG 0
1021 :    
1022 :     static __inline uint32_t Evaluate_Cost(const int16_t *C, int Mult, int Bias,
1023 :     const uint16_t * Zigzag, int Max, int Lambda)
1024 :     {
1025 :     #if (DBG>0)
1026 :     const int16_t * const Ref = C + 6*64;
1027 :     int Last = Max;
1028 :     int Bits = 0;
1029 :     int Dist = 0;
1030 :     int i;
1031 :     uint32_t Cost;
1032 :    
1033 :     while(Last>=0 && C[Zigzag[Last]]==0)
1034 :     Last--;
1035 :    
1036 :     if (Last>=0) {
1037 :     int j=0, j0=0;
1038 :     int Run, Level;
1039 :    
1040 :     Bits = 2; /* CBP */
1041 :     while(j<Last) {
1042 :     while(!C[Zigzag[j]])
1043 :     j++;
1044 :     if (j==Last)
1045 :     break;
1046 :     Level=C[Zigzag[j]];
1047 :     Run = j - j0;
1048 :     j0 = ++j;
1049 :     if (Level>=-24 && Level<=24)
1050 :     Bits += B16_17_Code_Len[(Level<0) ? -Level-1 : Level-1][Run];
1051 :     else
1052 :     Bits += 30;
1053 :     }
1054 :     Level = C[Zigzag[Last]];
1055 :     Run = j - j0;
1056 :     if (Level>=-6 && Level<=6)
1057 :     Bits += B16_17_Code_Len_Last[(Level<0) ? -Level-1 : Level-1][Run];
1058 :     else
1059 :     Bits += 30;
1060 :     }
1061 :    
1062 :     for(i=0; i<=Last; ++i) {
1063 :     int V = C[Zigzag[i]]*Mult;
1064 :     if (V>0)
1065 :     V += Bias;
1066 :     else
1067 :     if (V<0)
1068 :     V -= Bias;
1069 :     V -= Ref[Zigzag[i]];
1070 :     Dist += V*V;
1071 :     }
1072 :     Cost = Lambda*Dist + (Bits<<TL_SHIFT);
1073 :     if (DBG==1)
1074 :     printf( " Last:%2d/%2d Cost = [(Bits=%5.0d) + Lambda*(Dist=%6.0d) = %d ] >>12= %d ", Last,Max, Bits, Dist, Cost, Cost>>12 );
1075 :     return Cost;
1076 :    
1077 :     #else
1078 :     return 0;
1079 :     #endif
1080 :     }
1081 :    
1082 :    
1083 :     static int
1084 :     dct_quantize_trellis_h263_c(int16_t *const Out, const int16_t *const In, int Q, const uint16_t * const Zigzag, int Non_Zero)
1085 :     {
1086 :    
1087 :     /*
1088 :     * Note: We should search last non-zero coeffs on *real* DCT input coeffs (In[]),
1089 :     * not quantized one (Out[]). However, it only improves the result *very*
1090 :     * slightly (~0.01dB), whereas speed drops to crawling level :)
1091 :     * Well, actually, taking 1 more coeff past Non_Zero into account sometimes helps.
1092 :     */
1093 :     typedef struct { int16_t Run, Level; } NODE;
1094 :    
1095 :     NODE Nodes[65], Last;
1096 :     uint32_t Run_Costs0[64+1];
1097 :     uint32_t * const Run_Costs = Run_Costs0 + 1;
1098 :     const int Mult = 2*Q;
1099 :     const int Bias = (Q-1) | 1;
1100 :     const int Lev0 = Mult + Bias;
1101 :     const int Lambda = Trellis_Lambda_Tabs[Q-1]; /* it's 1/lambda, actually */
1102 :    
1103 :     int Run_Start = -1;
1104 :     Run_Costs[-1] = 2<<TL_SHIFT; /* source (w/ CBP penalty) */
1105 :     uint32_t Min_Cost = 2<<TL_SHIFT;
1106 :    
1107 :     int Last_Node = -1;
1108 :     uint32_t Last_Cost = 0;
1109 :    
1110 :     int i, j;
1111 :    
1112 :     #if (DBG>0)
1113 :     Last.Level = 0; Last.Run = -1; /* just initialize to smthg */
1114 :     #endif
1115 :    
1116 :     Non_Zero = Find_Last(Out, Zigzag, Non_Zero);
1117 :     if (Non_Zero<0)
1118 :     return -1;
1119 :    
1120 :     for(i=0; i<=Non_Zero; i++)
1121 :     {
1122 :     const int AC = In[Zigzag[i]];
1123 :     const int Level1 = Out[Zigzag[i]];
1124 :     const int Dist0 = Lambda* AC*AC;
1125 :     uint32_t Best_Cost = 0xf0000000;
1126 :     Last_Cost += Dist0;
1127 :    
1128 :     if ((uint32_t)(Level1+1)<3) /* very specialized loop for -1,0,+1 */
1129 :     {
1130 :     int dQ;
1131 :     int Run;
1132 :     uint32_t Cost0;
1133 :    
1134 :     if (AC<0) {
1135 :     Nodes[i].Level = -1;
1136 :     dQ = Lev0 + AC;
1137 :     } else {
1138 :     Nodes[i].Level = 1;
1139 :     dQ = Lev0 - AC;
1140 :     }
1141 :     Cost0 = Lambda*dQ*dQ;
1142 :    
1143 :     Nodes[i].Run = 1;
1144 :     Best_Cost = (Code_Len20[0]<<TL_SHIFT) + Run_Costs[i-1]+Cost0;
1145 :     for(Run=i-Run_Start; Run>0; --Run)
1146 :     {
1147 :     const uint32_t Cost_Base = Cost0 + Run_Costs[i-Run];
1148 :     const uint32_t Cost = Cost_Base + (Code_Len20[Run-1]<<TL_SHIFT);
1149 :     const uint32_t lCost = Cost_Base + (Code_Len24[Run-1]<<TL_SHIFT);
1150 :    
1151 :     /*
1152 :     * TODO: what about tie-breaks? Should we favor short runs or
1153 :     * long runs? Although the error is the same, it would not be
1154 :     * spread the same way along high and low frequencies...
1155 :     */
1156 :     if (Cost<Best_Cost) {
1157 :     Best_Cost = Cost;
1158 :     Nodes[i].Run = Run;
1159 :     }
1160 :    
1161 :     if (lCost<Last_Cost) {
1162 :     Last_Cost = lCost;
1163 :     Last.Run = Run;
1164 :     Last_Node = i;
1165 :     }
1166 :     }
1167 :     if (Last_Node==i)
1168 :     Last.Level = Nodes[i].Level;
1169 :    
1170 :     if (DBG==1) {
1171 :     Run_Costs[i] = Best_Cost;
1172 :     printf( "Costs #%2d: ", i);
1173 :     for(j=-1;j<=Non_Zero;++j) {
1174 :     if (j==Run_Start) printf( " %3.0d|", Run_Costs[j]>>12 );
1175 :     else if (j>Run_Start && j<i) printf( " %3.0d|", Run_Costs[j]>>12 );
1176 :     else if (j==i) printf( "(%3.0d)", Run_Costs[j]>>12 );
1177 :     else printf( " - |" );
1178 :     }
1179 :     printf( "<%3.0d %2d %d>", Min_Cost>>12, Nodes[i].Level, Nodes[i].Run );
1180 :     printf( " Last:#%2d {%3.0d %2d %d}", Last_Node, Last_Cost>>12, Last.Level, Last.Run );
1181 :     printf( " AC:%3.0d Dist0:%3d Dist(%d)=%d", AC, Dist0>>12, Nodes[i].Level, Cost0>>12 );
1182 :     printf( "\n" );
1183 :     }
1184 :     }
1185 :     else /* "big" levels */
1186 :     {
1187 :     const uint8_t *Tbl_L1, *Tbl_L2, *Tbl_L1_Last, *Tbl_L2_Last;
1188 :     int Level2;
1189 :     int dQ1, dQ2;
1190 :     int Run;
1191 :     uint32_t Dist1,Dist2;
1192 :     int dDist21;
1193 :    
1194 :     if (Level1>1) {
1195 :     dQ1 = Level1*Mult-AC + Bias;
1196 :     dQ2 = dQ1 - Mult;
1197 :     Level2 = Level1-1;
1198 :     Tbl_L1 = (Level1<=24) ? B16_17_Code_Len[Level1-1] : Code_Len0;
1199 :     Tbl_L2 = (Level2<=24) ? B16_17_Code_Len[Level2-1] : Code_Len0;
1200 :     Tbl_L1_Last = (Level1<=6) ? B16_17_Code_Len_Last[Level1-1] : Code_Len0;
1201 :     Tbl_L2_Last = (Level2<=6) ? B16_17_Code_Len_Last[Level2-1] : Code_Len0;
1202 :     } else { /* Level1<-1 */
1203 :     dQ1 = Level1*Mult-AC - Bias;
1204 :     dQ2 = dQ1 + Mult;
1205 :     Level2 = Level1 + 1;
1206 :     Tbl_L1 = (Level1>=-24) ? B16_17_Code_Len[Level1^-1] : Code_Len0;
1207 :     Tbl_L2 = (Level2>=-24) ? B16_17_Code_Len[Level2^-1] : Code_Len0;
1208 :     Tbl_L1_Last = (Level1>=- 6) ? B16_17_Code_Len_Last[Level1^-1] : Code_Len0;
1209 :     Tbl_L2_Last = (Level2>=- 6) ? B16_17_Code_Len_Last[Level2^-1] : Code_Len0;
1210 :     }
1211 :     Dist1 = Lambda*dQ1*dQ1;
1212 :     Dist2 = Lambda*dQ2*dQ2;
1213 :     dDist21 = Dist2-Dist1;
1214 :    
1215 :     for(Run=i-Run_Start; Run>0; --Run)
1216 :     {
1217 :     const uint32_t Cost_Base = Dist1 + Run_Costs[i-Run];
1218 :     uint32_t Cost1, Cost2;
1219 :     int bLevel;
1220 :    
1221 :     /*
1222 :     * for sub-optimal (but slightly worth it, speed-wise) search, uncomment the following:
1223 :     * if (Cost_Base>=Best_Cost) continue;
1224 :     */
1225 :     Cost1 = Cost_Base + (Tbl_L1[Run-1]<<TL_SHIFT);
1226 :     Cost2 = Cost_Base + (Tbl_L2[Run-1]<<TL_SHIFT) + dDist21;
1227 :    
1228 :     if (Cost2<Cost1) {
1229 :     Cost1 = Cost2;
1230 :     bLevel = Level2;
1231 :     } else
1232 :     bLevel = Level1;
1233 :    
1234 :     if (Cost1<Best_Cost) {
1235 :     Best_Cost = Cost1;
1236 :     Nodes[i].Run = Run;
1237 :     Nodes[i].Level = bLevel;
1238 :     }
1239 :    
1240 :     Cost1 = Cost_Base + (Tbl_L1_Last[Run-1]<<TL_SHIFT);
1241 :     Cost2 = Cost_Base + (Tbl_L2_Last[Run-1]<<TL_SHIFT) + dDist21;
1242 :    
1243 :     if (Cost2<Cost1) {
1244 :     Cost1 = Cost2;
1245 :     bLevel = Level2;
1246 :     } else
1247 :     bLevel = Level1;
1248 :    
1249 :     if (Cost1<Last_Cost) {
1250 :     Last_Cost = Cost1;
1251 :     Last.Run = Run;
1252 :     Last.Level = bLevel;
1253 :     Last_Node = i;
1254 :     }
1255 :     } /* end of "for Run" */
1256 :    
1257 :     if (DBG==1) {
1258 :     Run_Costs[i] = Best_Cost;
1259 :     printf( "Costs #%2d: ", i);
1260 :     for(j=-1;j<=Non_Zero;++j) {
1261 :     if (j==Run_Start) printf( " %3.0d|", Run_Costs[j]>>12 );
1262 :     else if (j>Run_Start && j<i) printf( " %3.0d|", Run_Costs[j]>>12 );
1263 :     else if (j==i) printf( "(%3.0d)", Run_Costs[j]>>12 );
1264 :     else printf( " - |" );
1265 :     }
1266 :     printf( "<%3.0d %2d %d>", Min_Cost>>12, Nodes[i].Level, Nodes[i].Run );
1267 :     printf( " Last:#%2d {%3.0d %2d %d}", Last_Node, Last_Cost>>12, Last.Level, Last.Run );
1268 :     printf( " AC:%3.0d Dist0:%3d Dist(%2d):%3d Dist(%2d):%3d", AC, Dist0>>12, Level1, Dist1>>12, Level2, Dist2>>12 );
1269 :     printf( "\n" );
1270 :     }
1271 :     }
1272 :    
1273 :     Run_Costs[i] = Best_Cost;
1274 :    
1275 :     if (Best_Cost < Min_Cost + Dist0) {
1276 :     Min_Cost = Best_Cost;
1277 :     Run_Start = i;
1278 :     }
1279 :     else
1280 :     {
1281 :     /*
1282 :     * as noticed by Michael Niedermayer (michaelni at gmx.at), there's
1283 :     * a code shorter by 1 bit for a larger run (!), same level. We give
1284 :     * it a chance by not moving the left barrier too much.
1285 :     */
1286 :    
1287 :     while( Run_Costs[Run_Start]>Min_Cost+(1<<TL_SHIFT) )
1288 :     Run_Start++;
1289 :    
1290 :     /* spread on preceding coeffs the cost incurred by skipping this one */
1291 :     for(j=Run_Start; j<i; ++j) Run_Costs[j] += Dist0;
1292 :     Min_Cost += Dist0;
1293 :     }
1294 :     }
1295 :    
1296 :     if (DBG) {
1297 :     Last_Cost = Evaluate_Cost(Out,Mult,Bias, Zigzag,Non_Zero, Lambda);
1298 :     if (DBG==1) {
1299 :     printf( "=> " );
1300 :     for(i=0; i<=Non_Zero; ++i) printf( "[%3.0d] ", Out[Zigzag[i]] );
1301 :     printf( "\n" );
1302 :     }
1303 :     }
1304 :    
1305 :     if (Last_Node<0)
1306 :     return -1;
1307 :    
1308 :     /* reconstruct optimal sequence backward with surviving paths */
1309 :     memset(Out, 0x00, 64*sizeof(*Out));
1310 :     Out[Zigzag[Last_Node]] = Last.Level;
1311 :     i = Last_Node - Last.Run;
1312 :     while(i>=0) {
1313 :     Out[Zigzag[i]] = Nodes[i].Level;
1314 :     i -= Nodes[i].Run;
1315 :     }
1316 :    
1317 :     if (DBG) {
1318 :     uint32_t Cost = Evaluate_Cost(Out,Mult,Bias, Zigzag,Non_Zero, Lambda);
1319 :     if (DBG==1) {
1320 :     printf( "<= " );
1321 :     for(i=0; i<=Last_Node; ++i) printf( "[%3.0d] ", Out[Zigzag[i]] );
1322 :     printf( "\n--------------------------------\n" );
1323 :     }
1324 :     if (Cost>Last_Cost) printf( "!!! %u > %u\n", Cost, Last_Cost );
1325 :     }
1326 :     return Last_Node;
1327 :     }
1328 :    
1329 :     #undef DBG
1330 :    
1331 :     #endif

No admin address has been configured
ViewVC Help
Powered by ViewVC 1.0.4