Parent Directory
|
Revision Log
Revision 1.7 - (view) (download)
1 : | Isibaar | 1.1 | /****************************************************************************** |
2 : | * * | ||
3 : | * This file is part of XviD, a free MPEG-4 video encoder/decoder * | ||
4 : | * * | ||
5 : | * XviD is an implementation of a part of one or more MPEG-4 Video tools * | ||
6 : | * as specified in ISO/IEC 14496-2 standard. Those intending to use this * | ||
7 : | * software module in hardware or software products are advised that its * | ||
8 : | * use may infringe existing patents or copyrights, and any such use * | ||
9 : | * would be at such party's own risk. The original developer of this * | ||
10 : | * software module and his/her company, and subsequent editors and their * | ||
11 : | * companies, will have no liability for use of this software or * | ||
12 : | * modifications or derivatives thereof. * | ||
13 : | * * | ||
14 : | * XviD is free software; you can redistribute it and/or modify it * | ||
15 : | * under the terms of the GNU General Public License as published by * | ||
16 : | * the Free Software Foundation; either version 2 of the License, or * | ||
17 : | * (at your option) any later version. * | ||
18 : | * * | ||
19 : | * XviD is distributed in the hope that it will be useful, but * | ||
20 : | * WITHOUT ANY WARRANTY; without even the implied warranty of * | ||
21 : | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * | ||
22 : | * GNU General Public License for more details. * | ||
23 : | * * | ||
24 : | * You should have received a copy of the GNU General Public License * | ||
25 : | * along with this program; if not, write to the Free Software * | ||
26 : | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * | ||
27 : | * * | ||
28 : | ******************************************************************************/ | ||
29 : | |||
30 : | /****************************************************************************** | ||
31 : | * * | ||
32 : | * mbtransquant.c * | ||
33 : | * * | ||
34 : | * Copyright (C) 2001 - Peter Ross <pross@cs.rmit.edu.au> * | ||
35 : | * Copyright (C) 2001 - Michael Militzer <isibaar@xvid.org> * | ||
36 : | * * | ||
37 : | * For more information visit the XviD homepage: http://www.xvid.org * | ||
38 : | * * | ||
39 : | ******************************************************************************/ | ||
40 : | |||
41 : | /****************************************************************************** | ||
42 : | * * | ||
43 : | * Revision history: * | ||
44 : | * * | ||
45 : | h | 1.4 | * 29.03.2002 interlacing speedup - used transfer strides instead of |
46 : | * manual field-to-frame conversion | ||
47 : | h | 1.2 | * 26.03.2002 interlacing support - moved transfers outside loops |
48 : | Isibaar | 1.1 | * 22.12.2001 get_dc_scaler() moved to common.h |
49 : | * 19.11.2001 introduced coefficient thresholding (Isibaar) * | ||
50 : | * 17.11.2001 initial version * | ||
51 : | * * | ||
52 : | ******************************************************************************/ | ||
53 : | |||
54 : | edgomez | 1.3 | #include <string.h> |
55 : | |||
56 : | Isibaar | 1.1 | #include "../portab.h" |
57 : | #include "mbfunctions.h" | ||
58 : | |||
59 : | #include "../global.h" | ||
60 : | #include "mem_transfer.h" | ||
61 : | #include "timer.h" | ||
62 : | #include "../dct/fdct.h" | ||
63 : | #include "../dct/idct.h" | ||
64 : | #include "../quant/quant_mpeg4.h" | ||
65 : | #include "../quant/quant_h263.h" | ||
66 : | #include "../encoder.h" | ||
67 : | |||
68 : | #define MIN(X, Y) ((X)<(Y)?(X):(Y)) | ||
69 : | #define MAX(X, Y) ((X)>(Y)?(X):(Y)) | ||
70 : | |||
71 : | edgomez | 1.7 | #define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ |
72 : | Isibaar | 1.1 | |
73 : | /* this isnt pretty, but its better than 20 ifdefs */ | ||
74 : | |||
75 : | edgomez | 1.7 | void |
76 : | MBTransQuantIntra(const MBParam * pParam, | ||
77 : | FRAMEINFO * frame, | ||
78 : | MACROBLOCK * pMB, | ||
79 : | const uint32_t x_pos, | ||
80 : | const uint32_t y_pos, | ||
81 : | int16_t data[6 * 64], | ||
82 : | int16_t qcoeff[6 * 64]) | ||
83 : | Isibaar | 1.1 | { |
84 : | edgomez | 1.3 | |
85 : | h | 1.4 | uint32_t stride = pParam->edged_width; |
86 : | uint32_t stride2 = stride / 2; | ||
87 : | uint32_t next_block = stride * 8; | ||
88 : | Isibaar | 1.1 | uint32_t i; |
89 : | suxen_drol | 1.6 | uint32_t iQuant = frame->quant; |
90 : | Isibaar | 1.1 | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
91 : | edgomez | 1.7 | IMAGE *pCurrent = &frame->image; |
92 : | Isibaar | 1.1 | |
93 : | edgomez | 1.3 | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
94 : | h | 1.4 | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
95 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
96 : | h | 1.2 | |
97 : | start_timer(); | ||
98 : | edgomez | 1.7 | transfer_8to16copy(&data[0 * 64], pY_Cur, stride); |
99 : | transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); | ||
100 : | transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); | ||
101 : | transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); | ||
102 : | transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); | ||
103 : | transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); | ||
104 : | h | 1.2 | stop_transfer_timer(); |
105 : | |||
106 : | start_timer(); | ||
107 : | pMB->field_dct = 0; | ||
108 : | edgomez | 1.7 | if ((frame->global_flags & XVID_INTERLACING)) { |
109 : | h | 1.2 | pMB->field_dct = MBDecideFieldDCT(data); |
110 : | } | ||
111 : | stop_interlacing_timer(); | ||
112 : | |||
113 : | edgomez | 1.7 | for (i = 0; i < 6; i++) { |
114 : | Isibaar | 1.1 | uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
115 : | |||
116 : | start_timer(); | ||
117 : | edgomez | 1.7 | fdct(&data[i * 64]); |
118 : | Isibaar | 1.1 | stop_dct_timer(); |
119 : | |||
120 : | edgomez | 1.7 | if (pParam->m_quant_type == H263_QUANT) { |
121 : | Isibaar | 1.1 | start_timer(); |
122 : | edgomez | 1.7 | quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
123 : | Isibaar | 1.1 | stop_quant_timer(); |
124 : | |||
125 : | start_timer(); | ||
126 : | edgomez | 1.7 | dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
127 : | Isibaar | 1.1 | stop_iquant_timer(); |
128 : | edgomez | 1.7 | } else { |
129 : | Isibaar | 1.1 | start_timer(); |
130 : | edgomez | 1.7 | quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
131 : | Isibaar | 1.1 | stop_quant_timer(); |
132 : | |||
133 : | start_timer(); | ||
134 : | edgomez | 1.7 | dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
135 : | Isibaar | 1.1 | stop_iquant_timer(); |
136 : | } | ||
137 : | |||
138 : | start_timer(); | ||
139 : | edgomez | 1.7 | idct(&data[i * 64]); |
140 : | Isibaar | 1.1 | stop_idct_timer(); |
141 : | edgomez | 1.3 | } |
142 : | h | 1.2 | |
143 : | edgomez | 1.7 | if (pMB->field_dct) { |
144 : | h | 1.4 | next_block = stride; |
145 : | stride *= 2; | ||
146 : | h | 1.2 | } |
147 : | Isibaar | 1.1 | |
148 : | h | 1.2 | start_timer(); |
149 : | edgomez | 1.7 | transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
150 : | transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); | ||
151 : | transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); | ||
152 : | transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride); | ||
153 : | transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); | ||
154 : | transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); | ||
155 : | h | 1.2 | stop_transfer_timer(); |
156 : | edgomez | 1.3 | |
157 : | Isibaar | 1.1 | } |
158 : | |||
159 : | |||
160 : | edgomez | 1.7 | uint8_t |
161 : | MBTransQuantInter(const MBParam * pParam, | ||
162 : | FRAMEINFO * frame, | ||
163 : | MACROBLOCK * pMB, | ||
164 : | const uint32_t x_pos, | ||
165 : | const uint32_t y_pos, | ||
166 : | int16_t data[6 * 64], | ||
167 : | int16_t qcoeff[6 * 64]) | ||
168 : | Isibaar | 1.1 | { |
169 : | edgomez | 1.3 | |
170 : | h | 1.4 | uint32_t stride = pParam->edged_width; |
171 : | uint32_t stride2 = stride / 2; | ||
172 : | uint32_t next_block = stride * 8; | ||
173 : | edgomez | 1.3 | uint32_t i; |
174 : | suxen_drol | 1.6 | uint32_t iQuant = frame->quant; |
175 : | Isibaar | 1.1 | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
176 : | edgomez | 1.3 | uint8_t cbp = 0; |
177 : | Isibaar | 1.1 | uint32_t sum; |
178 : | edgomez | 1.7 | IMAGE *pCurrent = &frame->image; |
179 : | |||
180 : | edgomez | 1.3 | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
181 : | h | 1.4 | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
182 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
183 : | Isibaar | 1.1 | |
184 : | h | 1.2 | start_timer(); |
185 : | pMB->field_dct = 0; | ||
186 : | edgomez | 1.7 | if ((frame->global_flags & XVID_INTERLACING)) { |
187 : | h | 1.2 | pMB->field_dct = MBDecideFieldDCT(data); |
188 : | } | ||
189 : | stop_interlacing_timer(); | ||
190 : | |||
191 : | edgomez | 1.7 | for (i = 0; i < 6; i++) { |
192 : | Isibaar | 1.1 | /* |
193 : | edgomez | 1.3 | * no need to transfer 8->16-bit |
194 : | * (this is performed already in motion compensation) | ||
195 : | */ | ||
196 : | Isibaar | 1.1 | start_timer(); |
197 : | edgomez | 1.7 | fdct(&data[i * 64]); |
198 : | Isibaar | 1.1 | stop_dct_timer(); |
199 : | |||
200 : | edgomez | 1.7 | if (pParam->m_quant_type == 0) { |
201 : | Isibaar | 1.1 | start_timer(); |
202 : | edgomez | 1.7 | sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
203 : | Isibaar | 1.1 | stop_quant_timer(); |
204 : | edgomez | 1.7 | } else { |
205 : | Isibaar | 1.1 | start_timer(); |
206 : | edgomez | 1.7 | sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
207 : | Isibaar | 1.1 | stop_quant_timer(); |
208 : | } | ||
209 : | |||
210 : | edgomez | 1.7 | if (sum >= TOOSMALL_LIMIT) { // skip block ? |
211 : | Isibaar | 1.1 | |
212 : | edgomez | 1.7 | if (pParam->m_quant_type == H263_QUANT) { |
213 : | Isibaar | 1.1 | start_timer(); |
214 : | edgomez | 1.7 | dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
215 : | Isibaar | 1.1 | stop_iquant_timer(); |
216 : | edgomez | 1.7 | } else { |
217 : | Isibaar | 1.1 | start_timer(); |
218 : | edgomez | 1.7 | dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
219 : | Isibaar | 1.1 | stop_iquant_timer(); |
220 : | } | ||
221 : | |||
222 : | cbp |= 1 << (5 - i); | ||
223 : | |||
224 : | start_timer(); | ||
225 : | edgomez | 1.7 | idct(&data[i * 64]); |
226 : | Isibaar | 1.1 | stop_idct_timer(); |
227 : | h | 1.2 | } |
228 : | } | ||
229 : | |||
230 : | edgomez | 1.7 | if (pMB->field_dct) { |
231 : | h | 1.4 | next_block = stride; |
232 : | stride *= 2; | ||
233 : | h | 1.2 | } |
234 : | |||
235 : | start_timer(); | ||
236 : | if (cbp & 32) | ||
237 : | edgomez | 1.7 | transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
238 : | h | 1.2 | if (cbp & 16) |
239 : | edgomez | 1.7 | transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
240 : | h | 1.2 | if (cbp & 8) |
241 : | edgomez | 1.7 | transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
242 : | h | 1.2 | if (cbp & 4) |
243 : | edgomez | 1.7 | transfer_16to8add(pY_Cur + next_block + 8, &data[3 * 64], stride); |
244 : | h | 1.2 | if (cbp & 2) |
245 : | edgomez | 1.7 | transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
246 : | h | 1.2 | if (cbp & 1) |
247 : | edgomez | 1.7 | transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
248 : | h | 1.2 | stop_transfer_timer(); |
249 : | |||
250 : | edgomez | 1.3 | return cbp; |
251 : | |||
252 : | h | 1.2 | } |
253 : | |||
254 : | |||
255 : | /* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ | ||
256 : | |||
257 : | Isibaar | 1.1 | |
258 : | edgomez | 1.7 | uint32_t |
259 : | MBDecideFieldDCT(int16_t data[6 * 64]) | ||
260 : | h | 1.2 | { |
261 : | edgomez | 1.3 | |
262 : | edgomez | 1.7 | const uint8_t blocks[] = |
263 : | { 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 }; | ||
264 : | const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 }; | ||
265 : | h | 1.2 | |
266 : | int frame = 0, field = 0; | ||
267 : | int i, j; | ||
268 : | |||
269 : | edgomez | 1.7 | for (i = 0; i < 7; ++i) { |
270 : | for (j = 0; j < 8; ++j) { | ||
271 : | frame += | ||
272 : | ABS(data[0 * 64 + (i + 1) * 8 + j] - data[0 * 64 + i * 8 + j]); | ||
273 : | frame += | ||
274 : | ABS(data[1 * 64 + (i + 1) * 8 + j] - data[1 * 64 + i * 8 + j]); | ||
275 : | frame += | ||
276 : | ABS(data[2 * 64 + (i + 1) * 8 + j] - data[2 * 64 + i * 8 + j]); | ||
277 : | frame += | ||
278 : | ABS(data[3 * 64 + (i + 1) * 8 + j] - data[3 * 64 + i * 8 + j]); | ||
279 : | |||
280 : | field += | ||
281 : | ABS(data[blocks[i + 1] + lines[i + 1] + j] - | ||
282 : | data[blocks[i] + lines[i] + j]); | ||
283 : | field += | ||
284 : | ABS(data[blocks[i + 1] + lines[i + 1] + 8 + j] - | ||
285 : | data[blocks[i] + lines[i] + 8 + j]); | ||
286 : | field += | ||
287 : | ABS(data[blocks[i + 1] + 64 + lines[i + 1] + j] - | ||
288 : | data[blocks[i] + 64 + lines[i] + j]); | ||
289 : | field += | ||
290 : | ABS(data[blocks[i + 1] + 64 + lines[i + 1] + 8 + j] - | ||
291 : | data[blocks[i] + 64 + lines[i] + 8 + j]); | ||
292 : | Isibaar | 1.1 | } |
293 : | } | ||
294 : | h | 1.2 | |
295 : | edgomez | 1.7 | if (frame > field) { |
296 : | h | 1.2 | MBFrameToField(data); |
297 : | } | ||
298 : | |||
299 : | return (frame > field); | ||
300 : | } | ||
301 : | |||
302 : | |||
303 : | /* deinterlace Y blocks vertically */ | ||
304 : | |||
305 : | #define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp)) | ||
306 : | edgomez | 1.3 | #define LINE(X,Y) &data[X*64 + Y*8] |
307 : | h | 1.2 | |
308 : | edgomez | 1.7 | void |
309 : | MBFrameToField(int16_t data[6 * 64]) | ||
310 : | h | 1.2 | { |
311 : | int16_t tmp[8]; | ||
312 : | |||
313 : | /* left blocks */ | ||
314 : | |||
315 : | // 1=2, 2=4, 4=8, 8=1 | ||
316 : | edgomez | 1.7 | MOVLINE(tmp, LINE(0, 1)); |
317 : | MOVLINE(LINE(0, 1), LINE(0, 2)); | ||
318 : | MOVLINE(LINE(0, 2), LINE(0, 4)); | ||
319 : | MOVLINE(LINE(0, 4), LINE(2, 0)); | ||
320 : | MOVLINE(LINE(2, 0), tmp); | ||
321 : | h | 1.2 | |
322 : | // 3=6, 6=12, 12=9, 9=3 | ||
323 : | edgomez | 1.7 | MOVLINE(tmp, LINE(0, 3)); |
324 : | MOVLINE(LINE(0, 3), LINE(0, 6)); | ||
325 : | MOVLINE(LINE(0, 6), LINE(2, 4)); | ||
326 : | MOVLINE(LINE(2, 4), LINE(2, 1)); | ||
327 : | MOVLINE(LINE(2, 1), tmp); | ||
328 : | h | 1.2 | |
329 : | // 5=10, 10=5 | ||
330 : | edgomez | 1.7 | MOVLINE(tmp, LINE(0, 5)); |
331 : | MOVLINE(LINE(0, 5), LINE(2, 2)); | ||
332 : | MOVLINE(LINE(2, 2), tmp); | ||
333 : | h | 1.2 | |
334 : | // 7=14, 14=13, 13=11, 11=7 | ||
335 : | edgomez | 1.7 | MOVLINE(tmp, LINE(0, 7)); |
336 : | MOVLINE(LINE(0, 7), LINE(2, 6)); | ||
337 : | MOVLINE(LINE(2, 6), LINE(2, 5)); | ||
338 : | MOVLINE(LINE(2, 5), LINE(2, 3)); | ||
339 : | MOVLINE(LINE(2, 3), tmp); | ||
340 : | h | 1.2 | |
341 : | /* right blocks */ | ||
342 : | |||
343 : | // 1=2, 2=4, 4=8, 8=1 | ||
344 : | edgomez | 1.7 | MOVLINE(tmp, LINE(1, 1)); |
345 : | MOVLINE(LINE(1, 1), LINE(1, 2)); | ||
346 : | MOVLINE(LINE(1, 2), LINE(1, 4)); | ||
347 : | MOVLINE(LINE(1, 4), LINE(3, 0)); | ||
348 : | MOVLINE(LINE(3, 0), tmp); | ||
349 : | h | 1.2 | |
350 : | // 3=6, 6=12, 12=9, 9=3 | ||
351 : | edgomez | 1.7 | MOVLINE(tmp, LINE(1, 3)); |
352 : | MOVLINE(LINE(1, 3), LINE(1, 6)); | ||
353 : | MOVLINE(LINE(1, 6), LINE(3, 4)); | ||
354 : | MOVLINE(LINE(3, 4), LINE(3, 1)); | ||
355 : | MOVLINE(LINE(3, 1), tmp); | ||
356 : | h | 1.2 | |
357 : | // 5=10, 10=5 | ||
358 : | edgomez | 1.7 | MOVLINE(tmp, LINE(1, 5)); |
359 : | MOVLINE(LINE(1, 5), LINE(3, 2)); | ||
360 : | MOVLINE(LINE(3, 2), tmp); | ||
361 : | h | 1.2 | |
362 : | // 7=14, 14=13, 13=11, 11=7 | ||
363 : | edgomez | 1.7 | MOVLINE(tmp, LINE(1, 7)); |
364 : | MOVLINE(LINE(1, 7), LINE(3, 6)); | ||
365 : | MOVLINE(LINE(3, 6), LINE(3, 5)); | ||
366 : | MOVLINE(LINE(3, 5), LINE(3, 3)); | ||
367 : | MOVLINE(LINE(3, 3), tmp); | ||
368 : | Isibaar | 1.1 | } |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |