Imported Upstream version 1.4
[deb_x265.git] / source / common / deblock.cpp
CommitLineData
72b9787e
JB
1/*****************************************************************************
2* Copyright (C) 2013 x265 project
3*
4* Author: Gopu Govindaswamy <gopu@multicorewareinc.com>
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 "deblock.h"
26#include "framedata.h"
27#include "picyuv.h"
28#include "slice.h"
29#include "mv.h"
30
31using namespace x265;
32
33#define DEBLOCK_SMALLEST_BLOCK 8
34#define DEFAULT_INTRA_TC_OFFSET 2
35
36void Deblock::deblockCTU(CUData* cu, int32_t dir)
37{
38 uint8_t blockingStrength[MAX_NUM_PARTITIONS];
39
40 memset(blockingStrength, 0, sizeof(uint8_t) * m_numPartitions);
41
42 deblockCU(cu, 0, 0, dir, blockingStrength);
43}
44
45/* Deblocking filter process in CU-based (the same function as conventional's)
46 * param Edge the direction of the edge in block boundary (horizonta/vertical), which is added newly */
47void Deblock::deblockCU(CUData* cu, uint32_t absPartIdx, uint32_t depth, const int32_t dir, uint8_t blockingStrength[])
48{
49 if (cu->m_partSize[absPartIdx] == SIZE_NONE)
50 return;
51
52 uint32_t curNumParts = NUM_CU_PARTITIONS >> (depth << 1);
53
54 const SPS& sps = *cu->m_slice->m_sps;
55
56 if (cu->m_cuDepth[absPartIdx] > depth)
57 {
58 uint32_t qNumParts = curNumParts >> 2;
59 uint32_t xmax = sps.picWidthInLumaSamples - cu->m_cuPelX;
60 uint32_t ymax = sps.picHeightInLumaSamples - cu->m_cuPelY;
61 for (uint32_t partIdx = 0; partIdx < 4; partIdx++, absPartIdx += qNumParts)
62 if (g_zscanToPelX[absPartIdx] < xmax && g_zscanToPelY[absPartIdx] < ymax)
63 deblockCU(cu, absPartIdx, depth + 1, dir, blockingStrength);
64 return;
65 }
66
67 const uint32_t widthInBaseUnits = sps.numPartInCUSize >> depth;
68 Param params;
69 setLoopfilterParam(cu, absPartIdx, &params);
70 setEdgefilterPU(cu, absPartIdx, dir, blockingStrength, widthInBaseUnits);
71 setEdgefilterTU(cu, absPartIdx, depth, dir, blockingStrength);
72 setEdgefilterMultiple(cu, absPartIdx, dir, 0, (dir == EDGE_VER ? params.leftEdge : params.topEdge), blockingStrength, widthInBaseUnits);
73
74 for (uint32_t partIdx = absPartIdx; partIdx < absPartIdx + curNumParts; partIdx++)
75 {
76 uint32_t bsCheck = !(partIdx & (1 << dir));
77
78 if (bsCheck && blockingStrength[partIdx])
79 getBoundaryStrengthSingle(cu, dir, partIdx, blockingStrength);
80 }
81
82 const uint32_t partIdxIncr = DEBLOCK_SMALLEST_BLOCK >> LOG2_UNIT_SIZE;
83 uint32_t sizeInPU = sps.numPartInCUSize >> depth;
84 uint32_t shiftFactor = (dir == EDGE_VER) ? cu->m_hChromaShift : cu->m_vChromaShift;
85 uint32_t chromaMask = ((DEBLOCK_SMALLEST_BLOCK << shiftFactor) >> LOG2_UNIT_SIZE) - 1;
86 uint32_t e0 = (dir == EDGE_VER ? g_zscanToPelX[absPartIdx] : g_zscanToPelY[absPartIdx]) >> LOG2_UNIT_SIZE;
87
88 for (uint32_t e = 0; e < sizeInPU; e += partIdxIncr)
89 {
90 edgeFilterLuma(cu, absPartIdx, depth, dir, e, blockingStrength);
91 if (!((e0 + e) & chromaMask))
92 edgeFilterChroma(cu, absPartIdx, depth, dir, e, blockingStrength);
93 }
94}
95
96static inline uint32_t calcBsIdx(CUData* cu, uint32_t absPartIdx, int32_t dir, int32_t edgeIdx, int32_t baseUnitIdx)
97{
98 uint32_t ctuWidthInBaseUnits = cu->m_slice->m_sps->numPartInCUSize;
99
100 if (dir)
101 return g_rasterToZscan[g_zscanToRaster[absPartIdx] + edgeIdx * ctuWidthInBaseUnits + baseUnitIdx];
102 else
103 return g_rasterToZscan[g_zscanToRaster[absPartIdx] + baseUnitIdx * ctuWidthInBaseUnits + edgeIdx];
104}
105
106void Deblock::setEdgefilterMultiple(CUData* cu, uint32_t scanIdx, int32_t dir, int32_t edgeIdx, uint8_t value, uint8_t blockingStrength[], uint32_t widthInBaseUnits)
107{
108 const uint32_t numElem = widthInBaseUnits;
109 X265_CHECK(numElem > 0, "numElem edge filter check\n");
110 for (uint32_t i = 0; i < numElem; i++)
111 {
112 const uint32_t bsidx = calcBsIdx(cu, scanIdx, dir, edgeIdx, i);
113 blockingStrength[bsidx] = value;
114 }
115}
116
117void Deblock::setEdgefilterTU(CUData* cu, uint32_t absPartIdx, uint32_t depth, int32_t dir, uint8_t blockingStrength[])
118{
119 if ((uint32_t)cu->m_tuDepth[absPartIdx] + cu->m_cuDepth[absPartIdx] > depth)
120 {
121 const uint32_t curNumParts = NUM_CU_PARTITIONS >> (depth << 1);
122 const uint32_t qNumParts = curNumParts >> 2;
123
124 for (uint32_t partIdx = 0; partIdx < 4; partIdx++, absPartIdx += qNumParts)
125 setEdgefilterTU(cu, absPartIdx, depth + 1, dir, blockingStrength);
126 return;
127 }
128
129 uint32_t widthInBaseUnits = 1 << (cu->m_log2CUSize[absPartIdx] - cu->m_tuDepth[absPartIdx] - LOG2_UNIT_SIZE);
130 setEdgefilterMultiple(cu, absPartIdx, dir, 0, 2, blockingStrength, widthInBaseUnits);
131}
132
133void Deblock::setEdgefilterPU(CUData* cu, uint32_t absPartIdx, int32_t dir, uint8_t blockingStrength[], uint32_t widthInBaseUnits)
134{
135 const uint32_t hWidthInBaseUnits = widthInBaseUnits >> 1;
136 const uint32_t qWidthInBaseUnits = widthInBaseUnits >> 2;
137
138 switch (cu->m_partSize[absPartIdx])
139 {
140 case SIZE_2NxN:
141 if (EDGE_HOR == dir)
142 setEdgefilterMultiple(cu, absPartIdx, dir, hWidthInBaseUnits, 1, blockingStrength, widthInBaseUnits);
143 break;
144 case SIZE_Nx2N:
145 if (EDGE_VER == dir)
146 setEdgefilterMultiple(cu, absPartIdx, dir, hWidthInBaseUnits, 1, blockingStrength, widthInBaseUnits);
147 break;
148 case SIZE_NxN:
149 setEdgefilterMultiple(cu, absPartIdx, dir, hWidthInBaseUnits, 1, blockingStrength, widthInBaseUnits);
150 break;
151 case SIZE_2NxnU:
152 if (EDGE_HOR == dir)
153 setEdgefilterMultiple(cu, absPartIdx, dir, qWidthInBaseUnits, 1, blockingStrength, widthInBaseUnits);
154 break;
155 case SIZE_nLx2N:
156 if (EDGE_VER == dir)
157 setEdgefilterMultiple(cu, absPartIdx, dir, qWidthInBaseUnits, 1, blockingStrength, widthInBaseUnits);
158 break;
159 case SIZE_2NxnD:
160 if (EDGE_HOR == dir)
161 setEdgefilterMultiple(cu, absPartIdx, dir, widthInBaseUnits - qWidthInBaseUnits, 1, blockingStrength, widthInBaseUnits);
162 break;
163 case SIZE_nRx2N:
164 if (EDGE_VER == dir)
165 setEdgefilterMultiple(cu, absPartIdx, dir, widthInBaseUnits - qWidthInBaseUnits, 1, blockingStrength, widthInBaseUnits);
166 break;
167
168 case SIZE_2Nx2N:
169 default:
170 break;
171 }
172}
173
174void Deblock::setLoopfilterParam(CUData* cu, uint32_t absPartIdx, Param *params)
175{
176 uint32_t x = cu->m_cuPelX + g_zscanToPelX[absPartIdx];
177 uint32_t y = cu->m_cuPelY + g_zscanToPelY[absPartIdx];
178
179 const CUData* tempCU;
180 uint32_t tempPartIdx;
181
182 if (!x)
183 params->leftEdge = 0;
184 else
185 {
186 tempCU = cu->getPULeft(tempPartIdx, absPartIdx);
187 if (tempCU)
188 params->leftEdge = 2;
189 else
190 params->leftEdge = 0;
191 }
192
193 if (!y)
194 params->topEdge = 0;
195 else
196 {
197 tempCU = cu->getPUAbove(tempPartIdx, absPartIdx);
198 if (tempCU)
199 params->topEdge = 2;
200 else
201 params->topEdge = 0;
202 }
203}
204
205void Deblock::getBoundaryStrengthSingle(CUData* cu, int32_t dir, uint32_t absPartIdx, uint8_t blockingStrength[])
206{
207 const Slice* const slice = cu->m_slice;
208 const uint32_t partQ = absPartIdx;
209 CUData* const cuQ = cu;
210
211 uint32_t partP;
212 const CUData* cuP;
213 uint8_t bs = 0;
214
215 // Calculate block index
216 if (dir == EDGE_VER)
217 cuP = cuQ->getPULeft(partP, partQ);
218 else // (dir == EDGE_HOR)
219 cuP = cuQ->getPUAbove(partP, partQ);
220
221 // Set BS for Intra MB : BS = 4 or 3
222 if (cuP->isIntra(partP) || cuQ->isIntra(partQ))
223 bs = 2;
224
225 // Set BS for not Intra MB : BS = 2 or 1 or 0
226 if (!cuP->isIntra(partP) && !cuQ->isIntra(partQ))
227 {
228 uint32_t nsPartQ = partQ;
229 uint32_t nsPartP = partP;
230
231 if (blockingStrength[absPartIdx] > 1 &&
232 (cuQ->getCbf(nsPartQ, TEXT_LUMA, cuQ->m_tuDepth[nsPartQ]) ||
233 cuP->getCbf(nsPartP, TEXT_LUMA, cuP->m_tuDepth[nsPartP])))
234 bs = 1;
235 else
236 {
237 if (dir == EDGE_HOR)
238 cuP = cuQ->getPUAbove(partP, partQ);
239
240 if (slice->isInterB() || cuP->m_slice->isInterB())
241 {
242 int32_t refIdx;
243 Frame *refP0, *refP1, *refQ0, *refQ1;
244 refIdx = cuP->m_refIdx[0][partP];
245 refP0 = (refIdx < 0) ? NULL : cuP->m_slice->m_refPicList[0][refIdx];
246 refIdx = cuP->m_refIdx[1][partP];
247 refP1 = (refIdx < 0) ? NULL : cuP->m_slice->m_refPicList[1][refIdx];
248 refIdx = cuQ->m_refIdx[0][partQ];
249 refQ0 = (refIdx < 0) ? NULL : slice->m_refPicList[0][refIdx];
250 refIdx = cuQ->m_refIdx[1][partQ];
251 refQ1 = (refIdx < 0) ? NULL : slice->m_refPicList[1][refIdx];
252
253 MV mvp0 = cuP->m_mv[0][partP];
254 MV mvp1 = cuP->m_mv[1][partP];
255 MV mvq0 = cuQ->m_mv[0][partQ];
256 MV mvq1 = cuQ->m_mv[1][partQ];
257
258 if (!refP0) mvp0 = 0;
259 if (!refP1) mvp1 = 0;
260 if (!refQ0) mvq0 = 0;
261 if (!refQ1) mvq1 = 0;
262
263 if (((refP0 == refQ0) && (refP1 == refQ1)) || ((refP0 == refQ1) && (refP1 == refQ0)))
264 {
265 if (refP0 != refP1) // Different L0 & L1
266 {
267 if (refP0 == refQ0)
268 {
269 bs = ((abs(mvq0.x - mvp0.x) >= 4) ||
270 (abs(mvq0.y - mvp0.y) >= 4) ||
271 (abs(mvq1.x - mvp1.x) >= 4) ||
272 (abs(mvq1.y - mvp1.y) >= 4)) ? 1 : 0;
273 }
274 else
275 {
276 bs = ((abs(mvq1.x - mvp0.x) >= 4) ||
277 (abs(mvq1.y - mvp0.y) >= 4) ||
278 (abs(mvq0.x - mvp1.x) >= 4) ||
279 (abs(mvq0.y - mvp1.y) >= 4)) ? 1 : 0;
280 }
281 }
282 else // Same L0 & L1
283 {
284 bs = ((abs(mvq0.x - mvp0.x) >= 4) ||
285 (abs(mvq0.y - mvp0.y) >= 4) ||
286 (abs(mvq1.x - mvp1.x) >= 4) ||
287 (abs(mvq1.y - mvp1.y) >= 4)) &&
288 ((abs(mvq1.x - mvp0.x) >= 4) ||
289 (abs(mvq1.y - mvp0.y) >= 4) ||
290 (abs(mvq0.x - mvp1.x) >= 4) ||
291 (abs(mvq0.y - mvp1.y) >= 4)) ? 1 : 0;
292 }
293 }
294 else // for all different Ref_Idx
295 bs = 1;
296 }
297 else // slice->isInterP()
298 {
299 int32_t refIdx;
300 Frame *refp0, *refq0;
301 refIdx = cuP->m_refIdx[0][partP];
302 refp0 = (refIdx < 0) ? NULL : cuP->m_slice->m_refPicList[0][refIdx];
303 refIdx = cuQ->m_refIdx[0][partQ];
304 refq0 = (refIdx < 0) ? NULL : slice->m_refPicList[0][refIdx];
305 MV mvp0 = cuP->m_mv[0][partP];
306 MV mvq0 = cuQ->m_mv[0][partQ];
307
308 if (!refp0) mvp0 = 0;
309 if (!refq0) mvq0 = 0;
310
311 bs = ((refp0 != refq0) ||
312 (abs(mvq0.x - mvp0.x) >= 4) ||
313 (abs(mvq0.y - mvp0.y) >= 4)) ? 1 : 0;
314 }
315 }
316 }
317
318 blockingStrength[absPartIdx] = bs;
319}
320
321static inline int32_t calcDP(pixel* src, intptr_t offset)
322{
323 return abs(static_cast<int32_t>(src[-offset * 3]) - 2 * src[-offset * 2] + src[-offset]);
324}
325
326static inline int32_t calcDQ(pixel* src, intptr_t offset)
327{
328 return abs(static_cast<int32_t>(src[0]) - 2 * src[offset] + src[offset * 2]);
329}
330
331static inline bool useStrongFiltering(intptr_t offset, int32_t beta, int32_t tc, pixel* src)
332{
333 int16_t m4 = (int16_t)src[0];
334 int16_t m3 = (int16_t)src[-offset];
335 int16_t m7 = (int16_t)src[offset * 3];
336 int16_t m0 = (int16_t)src[-offset * 4];
337 int32_t strong = abs(m0 - m3) + abs(m7 - m4);
338
339 return (strong < (beta >> 3)) && (abs(m3 - m4) < ((tc * 5 + 1) >> 1));
340}
341
342/* Deblocking for the luminance component with strong or weak filter
343 * \param src pointer to picture data
344 * \param offset offset value for picture data
345 * \param tc tc value
346 * \param partPNoFilter indicator to disable filtering on partP
347 * \param partQNoFilter indicator to disable filtering on partQ
348 * \param filterSecondP decision weak filter/no filter for partP
349 * \param filterSecondQ decision weak filter/no filter for partQ */
350static inline void pelFilterLumaStrong(pixel* src, intptr_t srcStep, intptr_t offset, int32_t tc, bool partPNoFilter, bool partQNoFilter)
351{
352 for (int32_t i = 0; i < UNIT_SIZE; i++, src += srcStep)
353 {
354 int16_t m4 = (int16_t)src[0];
355 int16_t m3 = (int16_t)src[-offset];
356 int16_t m5 = (int16_t)src[offset];
357 int16_t m2 = (int16_t)src[-offset * 2];
358 int32_t tc2 = 2 * tc;
359 if (!partPNoFilter)
360 {
361 int16_t m1 = (int16_t)src[-offset * 3];
362 int16_t m0 = (int16_t)src[-offset * 4];
363 src[-offset * 3] = (pixel)(Clip3(-tc2, tc2, ((2 * m0 + 3 * m1 + m2 + m3 + m4 + 4) >> 3) - m1) + m1);
364 src[-offset * 2] = (pixel)(Clip3(-tc2, tc2, ((m1 + m2 + m3 + m4 + 2) >> 2) - m2) + m2);
365 src[-offset] = (pixel)(Clip3(-tc2, tc2, ((m1 + 2 * m2 + 2 * m3 + 2 * m4 + m5 + 4) >> 3) - m3) + m3);
366 }
367 if (!partQNoFilter)
368 {
369 int16_t m6 = (int16_t)src[offset * 2];
370 int16_t m7 = (int16_t)src[offset * 3];
371 src[0] = (pixel)(Clip3(-tc2, tc2, ((m2 + 2 * m3 + 2 * m4 + 2 * m5 + m6 + 4) >> 3) - m4) + m4);
372 src[offset] = (pixel)(Clip3(-tc2, tc2, ((m3 + m4 + m5 + m6 + 2) >> 2) - m5) + m5);
373 src[offset * 2] = (pixel)(Clip3(-tc2, tc2, ((m3 + m4 + m5 + 3 * m6 + 2 * m7 + 4) >> 3) - m6) + m6);
374 }
375 }
376}
377
378/* Weak filter */
379static inline void pelFilterLuma(pixel* src, intptr_t srcStep, intptr_t offset, int32_t tc, bool partPNoFilter, bool partQNoFilter,
380 bool filterSecondP, bool filterSecondQ)
381{
382 int32_t thrCut = tc * 10;
383
384 for (int32_t i = 0; i < UNIT_SIZE; i++, src += srcStep)
385 {
386 int16_t m4 = (int16_t)src[0];
387 int16_t m3 = (int16_t)src[-offset];
388 int16_t m5 = (int16_t)src[offset];
389 int16_t m2 = (int16_t)src[-offset * 2];
390
391 int32_t delta = (9 * (m4 - m3) - 3 * (m5 - m2) + 8) >> 4;
392
393 if (abs(delta) < thrCut)
394 {
395 delta = Clip3(-tc, tc, delta);
396
397 int32_t tc2 = tc >> 1;
398 if (!partPNoFilter)
399 {
400 src[-offset] = Clip(m3 + delta);
401 if (filterSecondP)
402 {
403 int16_t m1 = (int16_t)src[-offset * 3];
404 int32_t delta1 = Clip3(-tc2, tc2, ((((m1 + m3 + 1) >> 1) - m2 + delta) >> 1));
405 src[-offset * 2] = Clip(m2 + delta1);
406 }
407 }
408 if (!partQNoFilter)
409 {
410 src[0] = Clip(m4 - delta);
411 if (filterSecondQ)
412 {
413 int16_t m6 = (int16_t)src[offset * 2];
414 int32_t delta2 = Clip3(-tc2, tc2, ((((m6 + m4 + 1) >> 1) - m5 - delta) >> 1));
415 src[offset] = Clip(m5 + delta2);
416 }
417 }
418 }
419 }
420}
421
422/* Deblocking of one line/column for the chrominance component
423 * \param src pointer to picture data
424 * \param offset offset value for picture data
425 * \param tc tc value
426 * \param partPNoFilter indicator to disable filtering on partP
427 * \param partQNoFilter indicator to disable filtering on partQ */
428static inline void pelFilterChroma(pixel* src, intptr_t srcStep, intptr_t offset, int32_t tc, bool partPNoFilter, bool partQNoFilter)
429{
430 for (int32_t i = 0; i < UNIT_SIZE; i++, src += srcStep)
431 {
432 int16_t m4 = (int16_t)src[0];
433 int16_t m3 = (int16_t)src[-offset];
434 int16_t m5 = (int16_t)src[offset];
435 int16_t m2 = (int16_t)src[-offset * 2];
436
437 int32_t delta = Clip3(-tc, tc, ((((m4 - m3) << 2) + m2 - m5 + 4) >> 3));
438 if (!partPNoFilter)
439 src[-offset] = Clip(m3 + delta);
440 if (!partQNoFilter)
441 src[0] = Clip(m4 - delta);
442 }
443}
444
445void Deblock::edgeFilterLuma(CUData* cu, uint32_t absPartIdx, uint32_t depth, int32_t dir, int32_t edge, const uint8_t blockingStrength[])
446{
447 PicYuv* reconYuv = cu->m_encData->m_reconPicYuv;
448 pixel* src = reconYuv->getLumaAddr(cu->m_cuAddr, absPartIdx);
449
450 intptr_t stride = reconYuv->m_stride;
451 uint32_t numParts = cu->m_slice->m_sps->numPartInCUSize >> depth;
452
453 intptr_t offset, srcStep;
454
455 bool partPNoFilter = false;
456 bool partQNoFilter = false;
457 uint32_t partP = 0;
458 uint32_t partQ = 0;
459 const CUData* cuP = cu;
460 const CUData* cuQ = cu;
461 int32_t betaOffset = cuQ->m_slice->m_pps->deblockingFilterBetaOffsetDiv2 << 1;
462 int32_t tcOffset = cuQ->m_slice->m_pps->deblockingFilterTcOffsetDiv2 << 1;
463
464 if (dir == EDGE_VER)
465 {
466 offset = 1;
467 srcStep = stride;
468 src += (edge << LOG2_UNIT_SIZE);
469 }
470 else // (dir == EDGE_HOR)
471 {
472 offset = stride;
473 srcStep = 1;
474 src += (edge << LOG2_UNIT_SIZE) * stride;
475 }
476
477 for (uint32_t idx = 0; idx < numParts; idx++)
478 {
479 uint32_t unitOffset = idx << LOG2_UNIT_SIZE;
480 uint32_t bsAbsIdx = calcBsIdx(cu, absPartIdx, dir, edge, idx);
481 uint32_t bs = blockingStrength[bsAbsIdx];
482 if (bs)
483 {
484 int32_t qpQ = cu->m_qp[bsAbsIdx];
485 partQ = bsAbsIdx;
486
487 // Derive neighboring PU index
488 if (dir == EDGE_VER)
489 cuP = cuQ->getPULeft(partP, partQ);
490 else // (dir == EDGE_HOR)
491 cuP = cuQ->getPUAbove(partP, partQ);
492
493 int32_t qpP = cuP->m_qp[partP];
494 int32_t qp = (qpP + qpQ + 1) >> 1;
495
496 int32_t indexB = Clip3(0, QP_MAX_SPEC, qp + betaOffset);
497
498 const int32_t bitdepthShift = X265_DEPTH - 8;
499 int32_t beta = s_betaTable[indexB] << bitdepthShift;
500
501 int32_t dp0 = calcDP(src + srcStep * (unitOffset + 0), offset);
502 int32_t dq0 = calcDQ(src + srcStep * (unitOffset + 0), offset);
503 int32_t dp3 = calcDP(src + srcStep * (unitOffset + 3), offset);
504 int32_t dq3 = calcDQ(src + srcStep * (unitOffset + 3), offset);
505 int32_t d0 = dp0 + dq0;
506 int32_t d3 = dp3 + dq3;
507
508 int32_t d = d0 + d3;
509
510 if (d < beta)
511 {
512 if (cu->m_slice->m_pps->bTransquantBypassEnabled)
513 {
514 // check if each of PUs is lossless coded
515 partPNoFilter = !!cuP->m_tqBypass[partP];
516 partQNoFilter = !!cuQ->m_tqBypass[partQ];
517 }
518
519 int32_t indexTC = Clip3(0, QP_MAX_SPEC + DEFAULT_INTRA_TC_OFFSET, int32_t(qp + DEFAULT_INTRA_TC_OFFSET * (bs - 1) + tcOffset));
520 int32_t tc = s_tcTable[indexTC] << bitdepthShift;
521
522 bool sw = (2 * d0 < (beta >> 2) &&
523 2 * d3 < (beta >> 2) &&
524 useStrongFiltering(offset, beta, tc, src + srcStep * (unitOffset + 0)) &&
525 useStrongFiltering(offset, beta, tc, src + srcStep * (unitOffset + 3)));
526
527 if (sw)
528 pelFilterLumaStrong(src + srcStep * unitOffset, srcStep, offset, tc, partPNoFilter, partQNoFilter);
529 else
530 {
531 int32_t sideThreshold = (beta + (beta >> 1)) >> 3;
532 int32_t dp = dp0 + dp3;
533 int32_t dq = dq0 + dq3;
534 bool filterP = (dp < sideThreshold);
535 bool filterQ = (dq < sideThreshold);
536
537 pelFilterLuma(src + srcStep * unitOffset, srcStep, offset, tc, partPNoFilter, partQNoFilter, filterP, filterQ);
538 }
539 }
540 }
541 }
542}
543
544void Deblock::edgeFilterChroma(CUData* cu, uint32_t absPartIdx, uint32_t depth, int32_t dir, int32_t edge, const uint8_t blockingStrength[])
545{
546 int32_t chFmt = cu->m_chromaFormat, chromaShift;
547 intptr_t offset, srcStep;
548
549 bool partPNoFilter = false;
550 bool partQNoFilter = false;
551 uint32_t partP;
552 uint32_t partQ;
553 const CUData* cuP;
554 const CUData* cuQ = cu;
555 int32_t tcOffset = cu->m_slice->m_pps->deblockingFilterTcOffsetDiv2 << 1;
556
557 X265_CHECK(((dir == EDGE_VER)
558 ? ((g_zscanToPelX[absPartIdx] + edge * UNIT_SIZE) >> cu->m_hChromaShift)
559 : ((g_zscanToPelY[absPartIdx] + edge * UNIT_SIZE) >> cu->m_vChromaShift)) % DEBLOCK_SMALLEST_BLOCK == 0,
560 "invalid edge\n");
561
562 PicYuv* reconPic = cu->m_encData->m_reconPicYuv;
563 intptr_t stride = reconPic->m_strideC;
564 intptr_t srcOffset = reconPic->getChromaAddrOffset(cu->m_cuAddr, absPartIdx);
565
566 if (dir == EDGE_VER)
567 {
568 chromaShift = cu->m_vChromaShift;
569 srcOffset += (edge << (LOG2_UNIT_SIZE - cu->m_hChromaShift));
570 offset = 1;
571 srcStep = stride;
572 }
573 else // (dir == EDGE_HOR)
574 {
575 chromaShift = cu->m_hChromaShift;
576 srcOffset += edge * stride << (LOG2_UNIT_SIZE - cu->m_vChromaShift);
577 offset = stride;
578 srcStep = 1;
579 }
580
581 pixel* srcChroma[2];
582 srcChroma[0] = reconPic->m_picOrg[1] + srcOffset;
583 srcChroma[1] = reconPic->m_picOrg[2] + srcOffset;
584
585 uint32_t numUnits = cu->m_slice->m_sps->numPartInCUSize >> (depth + chromaShift);
586
587 for (uint32_t idx = 0; idx < numUnits; idx++)
588 {
589 uint32_t unitOffset = idx << LOG2_UNIT_SIZE;
590 uint32_t bsAbsIdx = calcBsIdx(cu, absPartIdx, dir, edge, idx << chromaShift);
591 uint32_t bs = blockingStrength[bsAbsIdx];
592
593 if (bs > 1)
594 {
595 int32_t qpQ = cu->m_qp[bsAbsIdx];
596 partQ = bsAbsIdx;
597
598 // Derive neighboring PU index
599 if (dir == EDGE_VER)
600 cuP = cuQ->getPULeft(partP, partQ);
601 else // (dir == EDGE_HOR)
602 cuP = cuQ->getPUAbove(partP, partQ);
603
604 int32_t qpP = cuP->m_qp[partP];
605
606 if (cu->m_slice->m_pps->bTransquantBypassEnabled)
607 {
608 // check if each of PUs is lossless coded
609 partPNoFilter = !!cuP->m_tqBypass[partP];
610 partQNoFilter = !!cuQ->m_tqBypass[partQ];
611 }
612
613 for (uint32_t chromaIdx = 0; chromaIdx < 2; chromaIdx++)
614 {
615 int32_t chromaQPOffset = !chromaIdx ? cu->m_slice->m_pps->chromaCbQpOffset : cu->m_slice->m_pps->chromaCrQpOffset;
616 int32_t qp = ((qpP + qpQ + 1) >> 1) + chromaQPOffset;
617 if (qp >= 30)
618 {
619 if (chFmt == X265_CSP_I420)
620 qp = g_chromaScale[qp];
621 else
622 qp = X265_MIN(qp, 51);
623 }
624
625 int32_t indexTC = Clip3(0, QP_MAX_SPEC + DEFAULT_INTRA_TC_OFFSET, int32_t(qp + DEFAULT_INTRA_TC_OFFSET + tcOffset));
626 const int32_t bitdepthShift = X265_DEPTH - 8;
627 int32_t tc = s_tcTable[indexTC] << bitdepthShift;
628 pixel* srcC = srcChroma[chromaIdx];
629
630 pelFilterChroma(srcC + srcStep * unitOffset, srcStep, offset, tc, partPNoFilter, partQNoFilter);
631 }
632 }
633 }
634}
635
636const uint8_t Deblock::s_tcTable[54] =
637{
638 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2,
639 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 5, 5, 6, 6, 7, 8, 9, 10, 11, 13, 14, 16, 18, 20, 22, 24
640};
641
642const uint8_t Deblock::s_betaTable[52] =
643{
644 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17,
645 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64
646};
647