| 1 | /***************************************************************************** |
| 2 | * Copyright (C) 2014 x265 project |
| 3 | * |
| 4 | * Authors: Steve Borho <steve@borho.org> |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License as published by |
| 8 | * the Free Software Foundation; either version 2 of the License, or |
| 9 | * (at your option) any later version. |
| 10 | * |
| 11 | * This program is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU General Public License |
| 17 | * along with this program; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA. |
| 19 | * |
| 20 | * This program is also available under a commercial proprietary license. |
| 21 | * For more information, contact us at license @ x265.com. |
| 22 | *****************************************************************************/ |
| 23 | |
| 24 | #include "common.h" |
| 25 | #include "primitives.h" |
| 26 | #include "scalinglist.h" |
| 27 | |
| 28 | namespace { |
| 29 | // file-anonymous namespace |
| 30 | |
| 31 | /* Strings for scaling list file parsing */ |
| 32 | const char MatrixType[4][6][20] = |
| 33 | { |
| 34 | { |
| 35 | "INTRA4X4_LUMA", |
| 36 | "INTRA4X4_CHROMAU", |
| 37 | "INTRA4X4_CHROMAV", |
| 38 | "INTER4X4_LUMA", |
| 39 | "INTER4X4_CHROMAU", |
| 40 | "INTER4X4_CHROMAV" |
| 41 | }, |
| 42 | { |
| 43 | "INTRA8X8_LUMA", |
| 44 | "INTRA8X8_CHROMAU", |
| 45 | "INTRA8X8_CHROMAV", |
| 46 | "INTER8X8_LUMA", |
| 47 | "INTER8X8_CHROMAU", |
| 48 | "INTER8X8_CHROMAV" |
| 49 | }, |
| 50 | { |
| 51 | "INTRA16X16_LUMA", |
| 52 | "INTRA16X16_CHROMAU", |
| 53 | "INTRA16X16_CHROMAV", |
| 54 | "INTER16X16_LUMA", |
| 55 | "INTER16X16_CHROMAU", |
| 56 | "INTER16X16_CHROMAV" |
| 57 | }, |
| 58 | { |
| 59 | "INTRA32X32_LUMA", |
| 60 | "INTER32X32_LUMA", |
| 61 | }, |
| 62 | }; |
| 63 | const char MatrixType_DC[4][12][22] = |
| 64 | { |
| 65 | { |
| 66 | }, |
| 67 | { |
| 68 | }, |
| 69 | { |
| 70 | "INTRA16X16_LUMA_DC", |
| 71 | "INTRA16X16_CHROMAU_DC", |
| 72 | "INTRA16X16_CHROMAV_DC", |
| 73 | "INTER16X16_LUMA_DC", |
| 74 | "INTER16X16_CHROMAU_DC", |
| 75 | "INTER16X16_CHROMAV_DC" |
| 76 | }, |
| 77 | { |
| 78 | "INTRA32X32_LUMA_DC", |
| 79 | "INTER32X32_LUMA_DC", |
| 80 | }, |
| 81 | }; |
| 82 | |
| 83 | int quantTSDefault4x4[16] = |
| 84 | { |
| 85 | 16, 16, 16, 16, |
| 86 | 16, 16, 16, 16, |
| 87 | 16, 16, 16, 16, |
| 88 | 16, 16, 16, 16 |
| 89 | }; |
| 90 | |
| 91 | int quantIntraDefault8x8[64] = |
| 92 | { |
| 93 | 16, 16, 16, 16, 17, 18, 21, 24, |
| 94 | 16, 16, 16, 16, 17, 19, 22, 25, |
| 95 | 16, 16, 17, 18, 20, 22, 25, 29, |
| 96 | 16, 16, 18, 21, 24, 27, 31, 36, |
| 97 | 17, 17, 20, 24, 30, 35, 41, 47, |
| 98 | 18, 19, 22, 27, 35, 44, 54, 65, |
| 99 | 21, 22, 25, 31, 41, 54, 70, 88, |
| 100 | 24, 25, 29, 36, 47, 65, 88, 115 |
| 101 | }; |
| 102 | |
| 103 | int quantInterDefault8x8[64] = |
| 104 | { |
| 105 | 16, 16, 16, 16, 17, 18, 20, 24, |
| 106 | 16, 16, 16, 17, 18, 20, 24, 25, |
| 107 | 16, 16, 17, 18, 20, 24, 25, 28, |
| 108 | 16, 17, 18, 20, 24, 25, 28, 33, |
| 109 | 17, 18, 20, 24, 25, 28, 33, 41, |
| 110 | 18, 20, 24, 25, 28, 33, 41, 54, |
| 111 | 20, 24, 25, 28, 33, 41, 54, 71, |
| 112 | 24, 25, 28, 33, 41, 54, 71, 91 |
| 113 | }; |
| 114 | |
| 115 | } |
| 116 | |
| 117 | namespace x265 { |
| 118 | // private namespace |
| 119 | |
| 120 | const int ScalingList::s_numCoefPerSize[NUM_SIZES] = { 16, 64, 256, 1024 }; |
| 121 | const int32_t ScalingList::s_quantScales[NUM_REM] = { 26214, 23302, 20560, 18396, 16384, 14564 }; |
| 122 | const int32_t ScalingList::s_invQuantScales[NUM_REM] = { 40, 45, 51, 57, 64, 72 }; |
| 123 | |
| 124 | ScalingList::ScalingList() |
| 125 | { |
| 126 | memset(m_quantCoef, 0, sizeof(m_quantCoef)); |
| 127 | memset(m_dequantCoef, 0, sizeof(m_dequantCoef)); |
| 128 | memset(m_scalingListCoef, 0, sizeof(m_scalingListCoef)); |
| 129 | } |
| 130 | |
| 131 | bool ScalingList::init() |
| 132 | { |
| 133 | bool ok = true; |
| 134 | for (int sizeId = 0; sizeId < NUM_SIZES; sizeId++) |
| 135 | { |
| 136 | for (int listId = 0; listId < NUM_LISTS; listId++) |
| 137 | { |
| 138 | m_scalingListCoef[sizeId][listId] = X265_MALLOC(int32_t, X265_MIN(MAX_MATRIX_COEF_NUM, s_numCoefPerSize[sizeId])); |
| 139 | ok &= !!m_scalingListCoef[sizeId][listId]; |
| 140 | for (int rem = 0; rem < NUM_REM; rem++) |
| 141 | { |
| 142 | m_quantCoef[sizeId][listId][rem] = X265_MALLOC(int32_t, s_numCoefPerSize[sizeId]); |
| 143 | m_dequantCoef[sizeId][listId][rem] = X265_MALLOC(int32_t, s_numCoefPerSize[sizeId]); |
| 144 | ok &= m_quantCoef[sizeId][listId][rem] && m_dequantCoef[sizeId][listId][rem]; |
| 145 | } |
| 146 | } |
| 147 | } |
| 148 | return ok; |
| 149 | } |
| 150 | |
| 151 | ScalingList::~ScalingList() |
| 152 | { |
| 153 | for (int sizeId = 0; sizeId < NUM_SIZES; sizeId++) |
| 154 | { |
| 155 | for (int listId = 0; listId < NUM_LISTS; listId++) |
| 156 | { |
| 157 | X265_FREE(m_scalingListCoef[sizeId][listId]); |
| 158 | for (int rem = 0; rem < NUM_REM; rem++) |
| 159 | { |
| 160 | X265_FREE(m_quantCoef[sizeId][listId][rem]); |
| 161 | X265_FREE(m_dequantCoef[sizeId][listId][rem]); |
| 162 | } |
| 163 | } |
| 164 | } |
| 165 | } |
| 166 | |
| 167 | /* returns predicted list index if a match is found, else -1 */ |
| 168 | int ScalingList::checkPredMode(int size, int list) const |
| 169 | { |
| 170 | for (int predList = list; predList >= 0; predList--) |
| 171 | { |
| 172 | // check DC value |
| 173 | if (size < BLOCK_16x16 && m_scalingListDC[size][list] != m_scalingListDC[size][predList]) |
| 174 | continue; |
| 175 | |
| 176 | // check value of matrix |
| 177 | if (!memcmp(m_scalingListCoef[size][list], |
| 178 | list == predList ? getScalingListDefaultAddress(size, predList) : m_scalingListCoef[size][predList], |
| 179 | sizeof(int32_t) * X265_MIN(MAX_MATRIX_COEF_NUM, s_numCoefPerSize[size]))) |
| 180 | return predList; |
| 181 | } |
| 182 | |
| 183 | return -1; |
| 184 | } |
| 185 | |
| 186 | /* check if use default quantization matrix |
| 187 | * returns true if default quantization matrix is used in all sizes */ |
| 188 | bool ScalingList::checkDefaultScalingList() const |
| 189 | { |
| 190 | int defaultCounter = 0; |
| 191 | |
| 192 | for (int s = 0; s < NUM_SIZES; s++) |
| 193 | for (int l = 0; l < NUM_LISTS; l++) |
| 194 | if (!memcmp(m_scalingListCoef[s][l], getScalingListDefaultAddress(s, l), |
| 195 | sizeof(int32_t) * X265_MIN(MAX_MATRIX_COEF_NUM, s_numCoefPerSize[s])) && |
| 196 | ((s < BLOCK_16x16) || (m_scalingListDC[s][l] == 16))) |
| 197 | defaultCounter++; |
| 198 | |
| 199 | return defaultCounter != (NUM_LISTS * NUM_SIZES - 4); // -4 for 32x32 |
| 200 | } |
| 201 | |
| 202 | /* get address of default quantization matrix */ |
| 203 | const int32_t* ScalingList::getScalingListDefaultAddress(int sizeId, int listId) const |
| 204 | { |
| 205 | switch (sizeId) |
| 206 | { |
| 207 | case BLOCK_4x4: |
| 208 | return quantTSDefault4x4; |
| 209 | case BLOCK_8x8: |
| 210 | return (listId < 3) ? quantIntraDefault8x8 : quantInterDefault8x8; |
| 211 | case BLOCK_16x16: |
| 212 | return (listId < 3) ? quantIntraDefault8x8 : quantInterDefault8x8; |
| 213 | case BLOCK_32x32: |
| 214 | return (listId < 1) ? quantIntraDefault8x8 : quantInterDefault8x8; |
| 215 | default: |
| 216 | break; |
| 217 | } |
| 218 | |
| 219 | X265_CHECK(0, "invalid scaling list size\n"); |
| 220 | return NULL; |
| 221 | } |
| 222 | |
| 223 | void ScalingList::processDefaultMarix(int sizeId, int listId) |
| 224 | { |
| 225 | ::memcpy(m_scalingListCoef[sizeId][listId], getScalingListDefaultAddress(sizeId, listId), sizeof(int) * X265_MIN(MAX_MATRIX_COEF_NUM, s_numCoefPerSize[sizeId])); |
| 226 | m_scalingListDC[sizeId][listId] = SCALING_LIST_DC; |
| 227 | } |
| 228 | |
| 229 | void ScalingList::setDefaultScalingList() |
| 230 | { |
| 231 | for (int sizeId = 0; sizeId < NUM_SIZES; sizeId++) |
| 232 | for (int listId = 0; listId < NUM_LISTS; listId++) |
| 233 | processDefaultMarix(sizeId, listId); |
| 234 | m_bEnabled = true; |
| 235 | m_bDataPresent = false; |
| 236 | } |
| 237 | |
| 238 | bool ScalingList::parseScalingList(const char* filename) |
| 239 | { |
| 240 | FILE *fp = fopen(filename, "r"); |
| 241 | if (!fp) |
| 242 | { |
| 243 | x265_log(NULL, X265_LOG_ERROR, "can't open scaling list file %s\n", filename); |
| 244 | return true; |
| 245 | } |
| 246 | |
| 247 | char line[1024]; |
| 248 | int32_t *src = NULL; |
| 249 | |
| 250 | for (int sizeIdc = 0; sizeIdc < NUM_SIZES; sizeIdc++) |
| 251 | { |
| 252 | int size = X265_MIN(MAX_MATRIX_COEF_NUM, s_numCoefPerSize[sizeIdc]); |
| 253 | for (int listIdc = 0; listIdc < NUM_LISTS; listIdc++) |
| 254 | { |
| 255 | src = m_scalingListCoef[sizeIdc][listIdc]; |
| 256 | |
| 257 | fseek(fp, 0, 0); |
| 258 | do |
| 259 | { |
| 260 | char *ret = fgets(line, 1024, fp); |
| 261 | if (!ret || (!strstr(line, MatrixType[sizeIdc][listIdc]) && feof(fp))) |
| 262 | { |
| 263 | x265_log(NULL, X265_LOG_ERROR, "can't read matrix from %s\n", filename); |
| 264 | return true; |
| 265 | } |
| 266 | } |
| 267 | while (!strstr(line, MatrixType[sizeIdc][listIdc])); |
| 268 | |
| 269 | for (int i = 0; i < size; i++) |
| 270 | { |
| 271 | int data; |
| 272 | if (fscanf(fp, "%d,", &data) != 1) |
| 273 | { |
| 274 | x265_log(NULL, X265_LOG_ERROR, "can't read matrix from %s\n", filename); |
| 275 | return true; |
| 276 | } |
| 277 | src[i] = data; |
| 278 | } |
| 279 | |
| 280 | // set DC value for default matrix check |
| 281 | m_scalingListDC[sizeIdc][listIdc] = src[0]; |
| 282 | |
| 283 | if (sizeIdc > BLOCK_8x8) |
| 284 | { |
| 285 | fseek(fp, 0, 0); |
| 286 | do |
| 287 | { |
| 288 | char *ret = fgets(line, 1024, fp); |
| 289 | if (!ret || (!strstr(line, MatrixType_DC[sizeIdc][listIdc]) && feof(fp))) |
| 290 | { |
| 291 | x265_log(NULL, X265_LOG_ERROR, "can't read DC from %s\n", filename); |
| 292 | return true; |
| 293 | } |
| 294 | } |
| 295 | while (!strstr(line, MatrixType_DC[sizeIdc][listIdc])); |
| 296 | |
| 297 | int data; |
| 298 | if (fscanf(fp, "%d,", &data) != 1) |
| 299 | { |
| 300 | x265_log(NULL, X265_LOG_ERROR, "can't read matrix from %s\n", filename); |
| 301 | return true; |
| 302 | } |
| 303 | |
| 304 | // overwrite DC value when size of matrix is larger than 16x16 |
| 305 | m_scalingListDC[sizeIdc][listIdc] = data; |
| 306 | } |
| 307 | } |
| 308 | } |
| 309 | |
| 310 | fclose(fp); |
| 311 | |
| 312 | m_bEnabled = true; |
| 313 | m_bDataPresent = !checkDefaultScalingList(); |
| 314 | |
| 315 | return false; |
| 316 | } |
| 317 | |
| 318 | /** set quantized matrix coefficient for encode */ |
| 319 | void ScalingList::setupQuantMatrices() |
| 320 | { |
| 321 | for (int size = 0; size < NUM_SIZES; size++) |
| 322 | { |
| 323 | int width = 1 << (size + 2); |
| 324 | int ratio = width / X265_MIN(MAX_MATRIX_SIZE_NUM, width); |
| 325 | int stride = X265_MIN(MAX_MATRIX_SIZE_NUM, width); |
| 326 | int count = s_numCoefPerSize[size]; |
| 327 | |
| 328 | for (int list = 0; list < NUM_LISTS; list++) |
| 329 | { |
| 330 | int32_t *coeff = m_scalingListCoef[size][list]; |
| 331 | int32_t dc = m_scalingListDC[size][list]; |
| 332 | |
| 333 | for (int rem = 0; rem < NUM_REM; rem++) |
| 334 | { |
| 335 | int32_t *quantCoeff = m_quantCoef[size][list][rem]; |
| 336 | int32_t *dequantCoeff = m_dequantCoef[size][list][rem]; |
| 337 | |
| 338 | if (m_bEnabled) |
| 339 | { |
| 340 | processScalingListEnc(coeff, quantCoeff, s_quantScales[rem] << 4, width, width, ratio, stride, dc); |
| 341 | processScalingListDec(coeff, dequantCoeff, s_invQuantScales[rem], width, width, ratio, stride, dc); |
| 342 | } |
| 343 | else |
| 344 | { |
| 345 | /* flat quant and dequant coefficients */ |
| 346 | for (int i = 0; i < count; i++) |
| 347 | { |
| 348 | quantCoeff[i] = s_quantScales[rem]; |
| 349 | dequantCoeff[i] = s_invQuantScales[rem]; |
| 350 | } |
| 351 | } |
| 352 | } |
| 353 | } |
| 354 | } |
| 355 | } |
| 356 | |
| 357 | void ScalingList::processScalingListEnc(int32_t *coeff, int32_t *quantcoeff, int32_t quantScales, int height, int width, |
| 358 | int ratio, int stride, int32_t dc) |
| 359 | { |
| 360 | for (int j = 0; j < height; j++) |
| 361 | for (int i = 0; i < width; i++) |
| 362 | quantcoeff[j * width + i] = quantScales / coeff[stride * (j / ratio) + i / ratio]; |
| 363 | |
| 364 | if (ratio > 1) |
| 365 | quantcoeff[0] = quantScales / dc; |
| 366 | } |
| 367 | |
| 368 | void ScalingList::processScalingListDec(int32_t *coeff, int32_t *dequantcoeff, int32_t invQuantScales, int height, int width, |
| 369 | int ratio, int stride, int32_t dc) |
| 370 | { |
| 371 | for (int j = 0; j < height; j++) |
| 372 | for (int i = 0; i < width; i++) |
| 373 | dequantcoeff[j * width + i] = invQuantScales * coeff[stride * (j / ratio) + i / ratio]; |
| 374 | |
| 375 | if (ratio > 1) |
| 376 | dequantcoeff[0] = invQuantScales * dc; |
| 377 | } |
| 378 | |
| 379 | } |