Parent Directory
| 
Revision Log
Revision 82 - (view) (download)
| 1 : | Isibaar | 3 | /****************************************************************************** |
| 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 | 82 | * 29.03.2002 interlacing speedup - used transfer strides instead of |
| 46 : | * manual field-to-frame conversion | ||
| 47 : | h | 69 | * 26.03.2002 interlacing support - moved transfers outside loops |
| 48 : | Isibaar | 3 | * 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 | 78 | #include <string.h> |
| 55 : | |||
| 56 : | Isibaar | 3 | #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 : | #define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ | ||
| 72 : | |||
| 73 : | /* this isnt pretty, but its better than 20 ifdefs */ | ||
| 74 : | |||
| 75 : | void MBTransQuantIntra(const MBParam *pParam, | ||
| 76 : | edgomez | 78 | MACROBLOCK * pMB, |
| 77 : | Isibaar | 3 | const uint32_t x_pos, |
| 78 : | const uint32_t y_pos, | ||
| 79 : | edgomez | 78 | int16_t data[6*64], |
| 80 : | int16_t qcoeff[6*64], | ||
| 81 : | IMAGE * const pCurrent) | ||
| 82 : | Isibaar | 3 | |
| 83 : | { | ||
| 84 : | edgomez | 78 | |
| 85 : | h | 82 | uint32_t stride = pParam->edged_width; |
| 86 : | uint32_t stride2 = stride / 2; | ||
| 87 : | uint32_t next_block = stride * 8; | ||
| 88 : | Isibaar | 3 | uint32_t i; |
| 89 : | uint32_t iQuant = pParam->quant; | ||
| 90 : | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; | ||
| 91 : | |||
| 92 : | edgomez | 78 | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
| 93 : | h | 82 | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
| 94 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 95 : | h | 69 | |
| 96 : | start_timer(); | ||
| 97 : | h | 82 | transfer_8to16copy(&data[0*64], pY_Cur, stride); |
| 98 : | transfer_8to16copy(&data[1*64], pY_Cur + 8, stride); | ||
| 99 : | transfer_8to16copy(&data[2*64], pY_Cur + next_block, stride); | ||
| 100 : | transfer_8to16copy(&data[3*64], pY_Cur + next_block + 8,stride); | ||
| 101 : | transfer_8to16copy(&data[4*64], pU_Cur, stride2); | ||
| 102 : | transfer_8to16copy(&data[5*64], pV_Cur, stride2); | ||
| 103 : | h | 69 | stop_transfer_timer(); |
| 104 : | |||
| 105 : | start_timer(); | ||
| 106 : | pMB->field_dct = 0; | ||
| 107 : | if (pParam->global_flags & XVID_INTERLACING) | ||
| 108 : | { | ||
| 109 : | pMB->field_dct = MBDecideFieldDCT(data); | ||
| 110 : | } | ||
| 111 : | stop_interlacing_timer(); | ||
| 112 : | |||
| 113 : | for(i = 0; i < 6; i++) | ||
| 114 : | { | ||
| 115 : | Isibaar | 3 | uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
| 116 : | |||
| 117 : | start_timer(); | ||
| 118 : | edgomez | 78 | fdct(&data[i*64]); |
| 119 : | Isibaar | 3 | stop_dct_timer(); |
| 120 : | |||
| 121 : | if (pParam->quant_type == H263_QUANT) | ||
| 122 : | { | ||
| 123 : | start_timer(); | ||
| 124 : | edgomez | 78 | quant_intra(&qcoeff[i*64], &data[i*64], iQuant, iDcScaler); |
| 125 : | Isibaar | 3 | stop_quant_timer(); |
| 126 : | |||
| 127 : | start_timer(); | ||
| 128 : | edgomez | 78 | dequant_intra(&data[i*64], &qcoeff[i*64], iQuant, iDcScaler); |
| 129 : | Isibaar | 3 | stop_iquant_timer(); |
| 130 : | } | ||
| 131 : | else | ||
| 132 : | { | ||
| 133 : | start_timer(); | ||
| 134 : | edgomez | 78 | quant4_intra(&qcoeff[i*64], &data[i*64], iQuant, iDcScaler); |
| 135 : | Isibaar | 3 | stop_quant_timer(); |
| 136 : | |||
| 137 : | start_timer(); | ||
| 138 : | edgomez | 78 | dequant4_intra(&data[i*64], &qcoeff[i*64], iQuant, iDcScaler); |
| 139 : | Isibaar | 3 | stop_iquant_timer(); |
| 140 : | } | ||
| 141 : | |||
| 142 : | start_timer(); | ||
| 143 : | edgomez | 78 | idct(&data[i*64]); |
| 144 : | Isibaar | 3 | stop_idct_timer(); |
| 145 : | edgomez | 78 | } |
| 146 : | Isibaar | 3 | |
| 147 : | h | 69 | if (pMB->field_dct) |
| 148 : | { | ||
| 149 : | h | 82 | next_block = stride; |
| 150 : | stride *= 2; | ||
| 151 : | h | 69 | } |
| 152 : | |||
| 153 : | start_timer(); | ||
| 154 : | edgomez | 78 | transfer_16to8copy(pY_Cur, &data[0*64], stride); |
| 155 : | transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); | ||
| 156 : | h | 82 | transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); |
| 157 : | transfer_16to8copy(pY_Cur + next_block + 8, &data[3*64], stride); | ||
| 158 : | transfer_16to8copy(pU_Cur, &data[4*64], stride2); | ||
| 159 : | transfer_16to8copy(pV_Cur, &data[5*64], stride2); | ||
| 160 : | h | 69 | stop_transfer_timer(); |
| 161 : | edgomez | 78 | |
| 162 : | Isibaar | 3 | } |
| 163 : | |||
| 164 : | |||
| 165 : | uint8_t MBTransQuantInter(const MBParam *pParam, | ||
| 166 : | edgomez | 78 | MACROBLOCK * pMB, |
| 167 : | const uint32_t x_pos, const uint32_t y_pos, | ||
| 168 : | int16_t data[6*64], | ||
| 169 : | int16_t qcoeff[6*64], | ||
| 170 : | IMAGE * const pCurrent) | ||
| 171 : | Isibaar | 3 | |
| 172 : | { | ||
| 173 : | edgomez | 78 | |
| 174 : | h | 82 | uint32_t stride = pParam->edged_width; |
| 175 : | uint32_t stride2 = stride / 2; | ||
| 176 : | uint32_t next_block = stride * 8; | ||
| 177 : | edgomez | 78 | uint32_t i; |
| 178 : | uint32_t iQuant = pParam->quant; | ||
| 179 : | Isibaar | 3 | uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
| 180 : | edgomez | 78 | uint8_t cbp = 0; |
| 181 : | Isibaar | 3 | uint32_t sum; |
| 182 : | |||
| 183 : | edgomez | 78 | pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
| 184 : | h | 82 | pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
| 185 : | pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); | ||
| 186 : | Isibaar | 3 | |
| 187 : | h | 69 | start_timer(); |
| 188 : | pMB->field_dct = 0; | ||
| 189 : | if (pParam->global_flags & XVID_INTERLACING) | ||
| 190 : | { | ||
| 191 : | pMB->field_dct = MBDecideFieldDCT(data); | ||
| 192 : | } | ||
| 193 : | stop_interlacing_timer(); | ||
| 194 : | |||
| 195 : | edgomez | 78 | for(i = 0; i < 6; i++) |
| 196 : | h | 69 | { |
| 197 : | Isibaar | 3 | /* |
| 198 : | edgomez | 78 | * no need to transfer 8->16-bit |
| 199 : | * (this is performed already in motion compensation) | ||
| 200 : | */ | ||
| 201 : | Isibaar | 3 | start_timer(); |
| 202 : | edgomez | 78 | fdct(&data[i*64]); |
| 203 : | Isibaar | 3 | stop_dct_timer(); |
| 204 : | |||
| 205 : | if (pParam->quant_type == 0) | ||
| 206 : | { | ||
| 207 : | start_timer(); | ||
| 208 : | edgomez | 78 | sum = quant_inter(&qcoeff[i*64], &data[i*64], iQuant); |
| 209 : | Isibaar | 3 | stop_quant_timer(); |
| 210 : | } | ||
| 211 : | else | ||
| 212 : | { | ||
| 213 : | start_timer(); | ||
| 214 : | edgomez | 78 | sum = quant4_inter(&qcoeff[i*64], &data[i*64], iQuant); |
| 215 : | Isibaar | 3 | stop_quant_timer(); |
| 216 : | } | ||
| 217 : | |||
| 218 : | if(sum >= TOOSMALL_LIMIT) { // skip block ? | ||
| 219 : | |||
| 220 : | if (pParam->quant_type == H263_QUANT) | ||
| 221 : | { | ||
| 222 : | start_timer(); | ||
| 223 : | edgomez | 78 | dequant_inter(&data[i*64], &qcoeff[i*64], iQuant); |
| 224 : | Isibaar | 3 | stop_iquant_timer(); |
| 225 : | } | ||
| 226 : | else | ||
| 227 : | { | ||
| 228 : | start_timer(); | ||
| 229 : | edgomez | 78 | dequant4_inter(&data[i*64], &qcoeff[i*64], iQuant); |
| 230 : | Isibaar | 3 | stop_iquant_timer(); |
| 231 : | } | ||
| 232 : | |||
| 233 : | cbp |= 1 << (5 - i); | ||
| 234 : | |||
| 235 : | start_timer(); | ||
| 236 : | edgomez | 78 | idct(&data[i*64]); |
| 237 : | Isibaar | 3 | stop_idct_timer(); |
| 238 : | } | ||
| 239 : | } | ||
| 240 : | h | 69 | |
| 241 : | if (pMB->field_dct) | ||
| 242 : | { | ||
| 243 : | h | 82 | next_block = stride; |
| 244 : | stride *= 2; | ||
| 245 : | h | 69 | } |
| 246 : | |||
| 247 : | start_timer(); | ||
| 248 : | if (cbp & 32) | ||
| 249 : | edgomez | 78 | transfer_16to8add(pY_Cur, &data[0*64], stride); |
| 250 : | h | 69 | if (cbp & 16) |
| 251 : | edgomez | 78 | transfer_16to8add(pY_Cur + 8, &data[1*64], stride); |
| 252 : | h | 69 | if (cbp & 8) |
| 253 : | h | 82 | transfer_16to8add(pY_Cur + next_block, &data[2*64], stride); |
| 254 : | h | 69 | if (cbp & 4) |
| 255 : | h | 82 | transfer_16to8add(pY_Cur + next_block + 8, &data[3*64], stride); |
| 256 : | h | 69 | if (cbp & 2) |
| 257 : | h | 82 | transfer_16to8add(pU_Cur, &data[4*64], stride2); |
| 258 : | h | 69 | if (cbp & 1) |
| 259 : | h | 82 | transfer_16to8add(pV_Cur, &data[5*64], stride2); |
| 260 : | h | 69 | stop_transfer_timer(); |
| 261 : | |||
| 262 : | edgomez | 78 | return cbp; |
| 263 : | |||
| 264 : | Isibaar | 3 | } |
| 265 : | h | 69 | |
| 266 : | |||
| 267 : | /* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ | ||
| 268 : | |||
| 269 : | #define ABS(X) (X)<0 ? -(X) : (X) | ||
| 270 : | |||
| 271 : | edgomez | 78 | uint32_t MBDecideFieldDCT(int16_t data[6*64]) |
| 272 : | h | 69 | { |
| 273 : | |||
| 274 : | edgomez | 78 | const uint8_t blocks[] = {0*64, 0*64, 0*64, 0*64, 2*64, 2*64, 2*64, 2*64}; |
| 275 : | const uint8_t lines[] = {0, 16, 32, 48, 0, 16, 32, 48}; | ||
| 276 : | |||
| 277 : | h | 69 | int frame = 0, field = 0; |
| 278 : | int i, j; | ||
| 279 : | |||
| 280 : | for (i=0 ; i<7 ; ++i) | ||
| 281 : | { | ||
| 282 : | for (j=0 ; j<8 ; ++j) | ||
| 283 : | { | ||
| 284 : | edgomez | 78 | frame += ABS(data[0*64 + (i+1)*8 + j] - data[0*64 + i*8 + j]); |
| 285 : | frame += ABS(data[1*64 + (i+1)*8 + j] - data[1*64 + i*8 + j]); | ||
| 286 : | frame += ABS(data[2*64 + (i+1)*8 + j] - data[2*64 + i*8 + j]); | ||
| 287 : | frame += ABS(data[3*64 + (i+1)*8 + j] - data[3*64 + i*8 + j]); | ||
| 288 : | h | 69 | |
| 289 : | edgomez | 78 | field += ABS(data[blocks[i+1] + lines[i+1] + j] -\ |
| 290 : | data[blocks[i ] + lines[i ] + j]); | ||
| 291 : | field += ABS(data[blocks[i+1] + lines[i+1] + 8 + j] -\ | ||
| 292 : | data[blocks[i ] + lines[i ] + 8 + j]); | ||
| 293 : | field += ABS(data[blocks[i+1] + 64 + lines[i+1] + j] -\ | ||
| 294 : | data[blocks[i ] + 64 + lines[i ] + j]); | ||
| 295 : | field += ABS(data[blocks[i+1] + 64 + lines[i+1] + 8 + j] -\ | ||
| 296 : | data[blocks[i ] + 64 + lines[i ] + 8 + j]); | ||
| 297 : | h | 69 | } |
| 298 : | } | ||
| 299 : | |||
| 300 : | if (frame > field) | ||
| 301 : | { | ||
| 302 : | MBFrameToField(data); | ||
| 303 : | } | ||
| 304 : | |||
| 305 : | return (frame > field); | ||
| 306 : | } | ||
| 307 : | |||
| 308 : | |||
| 309 : | /* deinterlace Y blocks vertically */ | ||
| 310 : | |||
| 311 : | #define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp)) | ||
| 312 : | edgomez | 78 | #define LINE(X,Y) &data[X*64 + Y*8] |
| 313 : | h | 69 | |
| 314 : | edgomez | 78 | void MBFrameToField(int16_t data[6*64]) |
| 315 : | h | 69 | { |
| 316 : | int16_t tmp[8]; | ||
| 317 : | |||
| 318 : | /* left blocks */ | ||
| 319 : | |||
| 320 : | // 1=2, 2=4, 4=8, 8=1 | ||
| 321 : | MOVLINE(tmp, LINE(0,1)); | ||
| 322 : | MOVLINE(LINE(0,1), LINE(0,2)); | ||
| 323 : | MOVLINE(LINE(0,2), LINE(0,4)); | ||
| 324 : | MOVLINE(LINE(0,4), LINE(2,0)); | ||
| 325 : | MOVLINE(LINE(2,0), tmp); | ||
| 326 : | |||
| 327 : | // 3=6, 6=12, 12=9, 9=3 | ||
| 328 : | MOVLINE(tmp, LINE(0,3)); | ||
| 329 : | MOVLINE(LINE(0,3), LINE(0,6)); | ||
| 330 : | MOVLINE(LINE(0,6), LINE(2,4)); | ||
| 331 : | MOVLINE(LINE(2,4), LINE(2,1)); | ||
| 332 : | MOVLINE(LINE(2,1), tmp); | ||
| 333 : | |||
| 334 : | // 5=10, 10=5 | ||
| 335 : | MOVLINE(tmp, LINE(0,5)); | ||
| 336 : | MOVLINE(LINE(0,5), LINE(2,2)); | ||
| 337 : | MOVLINE(LINE(2,2), tmp); | ||
| 338 : | |||
| 339 : | // 7=14, 14=13, 13=11, 11=7 | ||
| 340 : | MOVLINE(tmp, LINE(0,7)); | ||
| 341 : | MOVLINE(LINE(0,7), LINE(2,6)); | ||
| 342 : | MOVLINE(LINE(2,6), LINE(2,5)); | ||
| 343 : | MOVLINE(LINE(2,5), LINE(2,3)); | ||
| 344 : | MOVLINE(LINE(2,3), tmp); | ||
| 345 : | |||
| 346 : | /* right blocks */ | ||
| 347 : | |||
| 348 : | // 1=2, 2=4, 4=8, 8=1 | ||
| 349 : | MOVLINE(tmp, LINE(1,1)); | ||
| 350 : | MOVLINE(LINE(1,1), LINE(1,2)); | ||
| 351 : | MOVLINE(LINE(1,2), LINE(1,4)); | ||
| 352 : | MOVLINE(LINE(1,4), LINE(3,0)); | ||
| 353 : | MOVLINE(LINE(3,0), tmp); | ||
| 354 : | |||
| 355 : | // 3=6, 6=12, 12=9, 9=3 | ||
| 356 : | MOVLINE(tmp, LINE(1,3)); | ||
| 357 : | MOVLINE(LINE(1,3), LINE(1,6)); | ||
| 358 : | MOVLINE(LINE(1,6), LINE(3,4)); | ||
| 359 : | MOVLINE(LINE(3,4), LINE(3,1)); | ||
| 360 : | MOVLINE(LINE(3,1), tmp); | ||
| 361 : | |||
| 362 : | // 5=10, 10=5 | ||
| 363 : | MOVLINE(tmp, LINE(1,5)); | ||
| 364 : | MOVLINE(LINE(1,5), LINE(3,2)); | ||
| 365 : | MOVLINE(LINE(3,2), tmp); | ||
| 366 : | |||
| 367 : | // 7=14, 14=13, 13=11, 11=7 | ||
| 368 : | MOVLINE(tmp, LINE(1,7)); | ||
| 369 : | MOVLINE(LINE(1,7), LINE(3,6)); | ||
| 370 : | MOVLINE(LINE(3,6), LINE(3,5)); | ||
| 371 : | MOVLINE(LINE(3,5), LINE(3,3)); | ||
| 372 : | MOVLINE(LINE(3,3), tmp); | ||
| 373 : | } |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |