Newer
Older

Karsten Suehring
committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/* The copyright in this software is being made available under the BSD
* License, included below. This software may be subject to other third party
* and contributor rights, including patent rights, and no such rights are
* granted under this license.
*
* Copyright (c) 2010-2018, ITU/ISO/IEC
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* * Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
* * Neither the name of the ITU/ISO/IEC nor the names of its contributors may
* be used to endorse or promote products derived from this software without
* specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
/** \file EncSlice.cpp
\brief slice encoder class
*/
#include "EncSlice.h"
#include "EncLib.h"
#include "CommonLib/UnitTools.h"
#include "CommonLib/Picture.h"
#if ENABLE_WPP_PARALLELISM
#include <mutex>
extern recursive_mutex g_cache_mutex;
#endif
#include <math.h>
//! \ingroup EncoderLib
//! \{
// ====================================================================================================================
// Constructor / destructor / create / destroy
// ====================================================================================================================
EncSlice::EncSlice()
: m_encCABACTableIdx(I_SLICE)
{
}
EncSlice::~EncSlice()
{
destroy();
}
void EncSlice::create( int iWidth, int iHeight, ChromaFormat chromaFormat, uint32_t iMaxCUWidth, uint32_t iMaxCUHeight, uint8_t uhTotalDepth )
{
}
void EncSlice::destroy()
{
// free lambda and QP arrays
m_vdRdPicLambda.clear();
m_vdRdPicQp.clear();
m_viRdPicQp.clear();
}
void EncSlice::init( EncLib* pcEncLib, const SPS& sps )
{
m_pcCfg = pcEncLib;
m_pcLib = pcEncLib;
m_pcListPic = pcEncLib->getListPic();
m_pcGOPEncoder = pcEncLib->getGOPEncoder();
m_pcCuEncoder = pcEncLib->getCuEncoder();
m_pcInterSearch = pcEncLib->getInterSearch();
m_CABACWriter = pcEncLib->getCABACEncoder()->getCABACWriter (&sps);
m_CABACEstimator = pcEncLib->getCABACEncoder()->getCABACEstimator(&sps);
m_pcTrQuant = pcEncLib->getTrQuant();
m_pcRdCost = pcEncLib->getRdCost();
// create lambda and QP arrays
m_vdRdPicLambda.resize(m_pcCfg->getDeltaQpRD() * 2 + 1 );
m_vdRdPicQp.resize( m_pcCfg->getDeltaQpRD() * 2 + 1 );
m_viRdPicQp.resize( m_pcCfg->getDeltaQpRD() * 2 + 1 );
m_pcRateCtrl = pcEncLib->getRateCtrl();
}
void
EncSlice::setUpLambda( Slice* slice, const double dLambda, int iQP)
{
// store lambda
m_pcRdCost ->setLambda( dLambda, slice->getSPS()->getBitDepths() );
// for RDO
// in RdCost there is only one lambda because the luma and chroma bits are not separated, instead we weight the distortion of chroma.
double dLambdas[MAX_NUM_COMPONENT] = { dLambda };
for( uint32_t compIdx = 1; compIdx < MAX_NUM_COMPONENT; compIdx++ )
{
const ComponentID compID = ComponentID( compIdx );
int chromaQPOffset = slice->getPPS()->getQpOffset( compID ) + slice->getSliceChromaQpDelta( compID );
int qpc = ( iQP + chromaQPOffset < 0 ) ? iQP : getScaledChromaQP( iQP + chromaQPOffset, m_pcCfg->getChromaFormatIdc() );
double tmpWeight = pow( 2.0, ( iQP - qpc ) / 3.0 ); // takes into account of the chroma qp mapping and chroma qp Offset
#if JVET_K0072
if( m_pcCfg->getDepQuantEnabledFlag() )
{
tmpWeight *= ( m_pcCfg->getGOPSize() >= 8 ? pow( 2.0, 0.1/3.0 ) : pow( 2.0, 0.2/3.0 ) ); // increase chroma weight for dependent quantization (in order to reduce bit rate shift from chroma to luma)
}
#endif
m_pcRdCost->setDistortionWeight( compID, tmpWeight );
#if ENABLE_WPP_PARALLELISM
for( int jId = 1; jId < ( m_pcLib->getNumWppThreads() + m_pcLib->getNumWppExtraLines() ); jId++ )
{
m_pcLib->getRdCost( slice->getPic()->scheduler.getWppDataId( jId ) )->setDistortionWeight( compID, tmpWeight );
}
#endif
dLambdas[compIdx] = dLambda / tmpWeight;
}
#if RDOQ_CHROMA_LAMBDA
// for RDOQ
m_pcTrQuant->setLambdas( dLambdas );
#else
m_pcTrQuant->setLambda( dLambda );
#endif
// for SAO
slice->setLambdas( dLambdas );
}
/**
- non-referenced frame marking
- QP computation based on temporal structure
- lambda computation based on QP
- set temporal layer ID and the parameter sets
.
\param pcPic picture class
\param pocLast POC of last picture
\param pocCurr current POC
\param iNumPicRcvd number of received pictures
\param iGOPid POC offset for hierarchical structure
\param rpcSlice slice header class
\param isField true for field coding
*/
void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCurr, const int iGOPid, Slice*& rpcSlice, const bool isField )
{
double dQP;
double dLambda;
rpcSlice = pcPic->slices[0];
rpcSlice->setSliceBits(0);
rpcSlice->setPic( pcPic );
rpcSlice->initSlice();
rpcSlice->setPicOutputFlag( true );
rpcSlice->setPOC( pocCurr );
#if JVET_K0072
rpcSlice->setDepQuantEnabledFlag( m_pcCfg->getDepQuantEnabledFlag() );
#if HEVC_USE_SIGN_HIDING
rpcSlice->setSignDataHidingEnabledFlag( m_pcCfg->getSignDataHidingEnabledFlag() );
#endif
#endif
#if SHARP_LUMA_DELTA_QP
pcPic->fieldPic = isField;
m_gopID = iGOPid;
#endif
// depth computation based on GOP size
int depth;
{
int poc = rpcSlice->getPOC();
if(isField)
{
poc = (poc/2) % (m_pcCfg->getGOPSize()/2);
}
else
{
poc = poc % m_pcCfg->getGOPSize();
}
if ( poc == 0 )
{
depth = 0;
}
else
{
int step = m_pcCfg->getGOPSize();
depth = 0;
for( int i=step>>1; i>=1; i>>=1 )
{
for ( int j=i; j<m_pcCfg->getGOPSize(); j+=step )
{
if ( j == poc )
{
i=0;
break;
}
}
step >>= 1;
depth++;
}
}
if(m_pcCfg->getHarmonizeGopFirstFieldCoupleEnabled() && poc != 0)
{
if (isField && ((rpcSlice->getPOC() % 2) == 1))
{
depth++;
}
}
}
// slice type
SliceType eSliceType;
eSliceType=B_SLICE;
if(!(isField && pocLast == 1) || !m_pcCfg->getEfficientFieldIRAPEnabled())
{
if(m_pcCfg->getDecodingRefreshType() == 3)
{
eSliceType = (pocLast == 0 || pocCurr % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
}
else
{
eSliceType = (pocLast == 0 || (pocCurr - (isField ? 1 : 0)) % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
}
}
rpcSlice->setSliceType ( eSliceType );
// ------------------------------------------------------------------------------------------------------------------
// Non-referenced frame marking
// ------------------------------------------------------------------------------------------------------------------
if(pocLast == 0)
{
rpcSlice->setTemporalLayerNonReferenceFlag(false);
}
else
{
rpcSlice->setTemporalLayerNonReferenceFlag(!m_pcCfg->getGOPEntry(iGOPid).m_refPic);
}
pcPic->referenced = true;
// ------------------------------------------------------------------------------------------------------------------
// QP setting
// ------------------------------------------------------------------------------------------------------------------
#if X0038_LAMBDA_FROM_QP_CAPABILITY
dQP = m_pcCfg->getQPForPicture(iGOPid, rpcSlice);
#else
dQP = m_pcCfg->getBaseQP();
if(eSliceType!=I_SLICE)
{
#if SHARP_LUMA_DELTA_QP
if (!(( m_pcCfg->getMaxDeltaQP() == 0) && (!m_pcCfg->getLumaLevelToDeltaQPMapping().isEnabled()) && (dQP == -rpcSlice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA) ) && (rpcSlice->getPPS()->getTransquantBypassEnabledFlag())))
#else
if (!(( m_pcCfg->getMaxDeltaQP() == 0 ) && (dQP == -rpcSlice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA) ) && (rpcSlice->getPPS()->getTransquantBypassEnabledFlag())))
#endif
{
dQP += m_pcCfg->getGOPEntry(iGOPid).m_QPOffset;
}
}
// modify QP
const int* pdQPs = m_pcCfg->getdQPs();
if ( pdQPs )
{
dQP += pdQPs[ rpcSlice->getPOC() ];
}
if (m_pcCfg->getCostMode()==COST_LOSSLESS_CODING)
{
dQP=LOSSLESS_AND_MIXED_LOSSLESS_RD_COST_TEST_QP;
m_pcCfg->setDeltaQpRD(0);
}
#endif
// ------------------------------------------------------------------------------------------------------------------
// Lambda computation
// ------------------------------------------------------------------------------------------------------------------
#if X0038_LAMBDA_FROM_QP_CAPABILITY
const int temporalId=m_pcCfg->getGOPEntry(iGOPid).m_temporalId;
#if !SHARP_LUMA_DELTA_QP
const std::vector<double> &intraLambdaModifiers=m_pcCfg->getIntraLambdaModifier();
#endif
#endif
int iQP;
double dOrigQP = dQP;
// pre-compute lambda and QP values for all possible QP candidates
for ( int iDQpIdx = 0; iDQpIdx < 2 * m_pcCfg->getDeltaQpRD() + 1; iDQpIdx++ )
{
// compute QP value
dQP = dOrigQP + ((iDQpIdx+1)>>1)*(iDQpIdx%2 ? -1 : 1);
#if SHARP_LUMA_DELTA_QP
dLambda = calculateLambda(rpcSlice, iGOPid, depth, dQP, dQP, iQP );
#else
// compute lambda value
int NumberBFrames = ( m_pcCfg->getGOPSize() - 1 );
int SHIFT_QP = 12;
#if DISTORTION_LAMBDA_BUGFIX
int bitdepth_luma_qp_scale =
6
* (rpcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8
- DISTORTION_PRECISION_ADJUSTMENT(rpcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA)));
#else
#if FULL_NBIT
int bitdepth_luma_qp_scale = 6 * (rpcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8);
#else
int bitdepth_luma_qp_scale = 0;
#endif
#endif
double qp_temp = (double) dQP + bitdepth_luma_qp_scale - SHIFT_QP;
#if FULL_NBIT
double qp_temp_orig = (double) dQP - SHIFT_QP;
#endif
// Case #1: I or P-slices (key-frame)
double dQPFactor = m_pcCfg->getGOPEntry(iGOPid).m_QPFactor;
if ( eSliceType==I_SLICE )
{
if (m_pcCfg->getIntraQpFactor()>=0.0 && m_pcCfg->getGOPEntry(iGOPid).m_sliceType != I_SLICE)
{
dQPFactor=m_pcCfg->getIntraQpFactor();
}
else
{
#if X0038_LAMBDA_FROM_QP_CAPABILITY
if(m_pcCfg->getLambdaFromQPEnable())
{
dQPFactor=0.57;
}
else
{
#endif
double dLambda_scale = 1.0 - Clip3( 0.0, 0.5, 0.05*(double)(isField ? NumberBFrames/2 : NumberBFrames) );
dQPFactor=0.57*dLambda_scale;
#if X0038_LAMBDA_FROM_QP_CAPABILITY
}
#endif
}
}
#if X0038_LAMBDA_FROM_QP_CAPABILITY
else if( m_pcCfg->getLambdaFromQPEnable() )
{
dQPFactor=0.57;
}
#endif
dLambda = dQPFactor*pow( 2.0, qp_temp/3.0 );
#if X0038_LAMBDA_FROM_QP_CAPABILITY
if(!m_pcCfg->getLambdaFromQPEnable() && depth>0)
#else
if ( depth>0 )
#endif
{
#if FULL_NBIT
dLambda *= Clip3( 2.00, 4.00, (qp_temp_orig / 6.0) ); // (j == B_SLICE && p_cur_frm->layer != 0 )
#else
dLambda *= Clip3( 2.00, 4.00, (qp_temp / 6.0) ); // (j == B_SLICE && p_cur_frm->layer != 0 )
#endif
}
// if hadamard is used in ME process
if ( !m_pcCfg->getUseHADME() && rpcSlice->getSliceType( ) != I_SLICE )
{
dLambda *= 0.95;
}
#if X0038_LAMBDA_FROM_QP_CAPABILITY
double lambdaModifier;
if( rpcSlice->getSliceType( ) != I_SLICE || intraLambdaModifiers.empty())
{
lambdaModifier = m_pcCfg->getLambdaModifier( temporalId );
}
else
{
lambdaModifier = intraLambdaModifiers[ (temporalId < intraLambdaModifiers.size()) ? temporalId : (intraLambdaModifiers.size()-1) ];
}
dLambda *= lambdaModifier;
#endif
iQP = max( -rpcSlice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA), min( MAX_QP, (int) floor( dQP + 0.5 ) ) );
#endif
m_vdRdPicLambda[iDQpIdx] = dLambda;
m_vdRdPicQp [iDQpIdx] = dQP;
m_viRdPicQp [iDQpIdx] = iQP;
}
// obtain dQP = 0 case
dLambda = m_vdRdPicLambda[0];
dQP = m_vdRdPicQp [0];
iQP = m_viRdPicQp [0];
#if !X0038_LAMBDA_FROM_QP_CAPABILITY
const int temporalId=m_pcCfg->getGOPEntry(iGOPid).m_temporalId;
const std::vector<double> &intraLambdaModifiers=m_pcCfg->getIntraLambdaModifier();
#endif
#if W0038_CQP_ADJ
if(rpcSlice->getPPS()->getSliceChromaQpFlag())
{
const bool bUseIntraOrPeriodicOffset = rpcSlice->getSliceType()==I_SLICE || (m_pcCfg->getSliceChromaOffsetQpPeriodicity()!=0 && (rpcSlice->getPOC()%m_pcCfg->getSliceChromaOffsetQpPeriodicity())==0);
int cbQP = bUseIntraOrPeriodicOffset? m_pcCfg->getSliceChromaOffsetQpIntraOrPeriodic(false) : m_pcCfg->getGOPEntry(iGOPid).m_CbQPoffset;
int crQP = bUseIntraOrPeriodicOffset? m_pcCfg->getSliceChromaOffsetQpIntraOrPeriodic(true) : m_pcCfg->getGOPEntry(iGOPid).m_CrQPoffset;
cbQP = Clip3( -12, 12, cbQP + rpcSlice->getPPS()->getQpOffset(COMPONENT_Cb) ) - rpcSlice->getPPS()->getQpOffset(COMPONENT_Cb);
crQP = Clip3( -12, 12, crQP + rpcSlice->getPPS()->getQpOffset(COMPONENT_Cr) ) - rpcSlice->getPPS()->getQpOffset(COMPONENT_Cr);
rpcSlice->setSliceChromaQpDelta(COMPONENT_Cb, Clip3( -12, 12, cbQP));
CHECK(!(rpcSlice->getSliceChromaQpDelta(COMPONENT_Cb)+rpcSlice->getPPS()->getQpOffset(COMPONENT_Cb)<=12 && rpcSlice->getSliceChromaQpDelta(COMPONENT_Cb)+rpcSlice->getPPS()->getQpOffset(COMPONENT_Cb)>=-12), "Unspecified error");
rpcSlice->setSliceChromaQpDelta(COMPONENT_Cr, Clip3( -12, 12, crQP));
CHECK(!(rpcSlice->getSliceChromaQpDelta(COMPONENT_Cr)+rpcSlice->getPPS()->getQpOffset(COMPONENT_Cr)<=12 && rpcSlice->getSliceChromaQpDelta(COMPONENT_Cr)+rpcSlice->getPPS()->getQpOffset(COMPONENT_Cr)>=-12), "Unspecified error");
}
else
{
rpcSlice->setSliceChromaQpDelta( COMPONENT_Cb, 0 );
rpcSlice->setSliceChromaQpDelta( COMPONENT_Cr, 0 );
}
#endif
#if !X0038_LAMBDA_FROM_QP_CAPABILITY
double lambdaModifier;
if( rpcSlice->getSliceType( ) != I_SLICE || intraLambdaModifiers.empty())
{
lambdaModifier = m_pcCfg->getLambdaModifier( temporalId );
}
else
{
lambdaModifier = intraLambdaModifiers[ (temporalId < intraLambdaModifiers.size()) ? temporalId : (intraLambdaModifiers.size()-1) ];
}
dLambda *= lambdaModifier;
#endif
setUpLambda(rpcSlice, dLambda, iQP);
#if WCG_EXT
// cost = Distortion + Lambda*R,
// when QP is adjusted by luma, distortion is changed, so we have to adjust lambda to match the distortion, then the cost function becomes
// costA = Distortion + AdjustedLambda * R -- currently, costA is still used when calculating intermediate cost of using SAD, HAD, resisual etc.
// an alternative way is to weight the distortion to before the luma QP adjustment, then the cost function becomes
// costB = weightedDistortion + Lambda * R -- currently, costB is used to calculat final cost, and when DF_FUNC is DF_DEFAULT
m_pcRdCost->saveUnadjustedLambda();
#endif
if (m_pcCfg->getFastMEForGenBLowDelayEnabled())
{
// restore original slice type
if(!(isField && pocLast == 1) || !m_pcCfg->getEfficientFieldIRAPEnabled())
{
if(m_pcCfg->getDecodingRefreshType() == 3)
{
eSliceType = (pocLast == 0 || (pocCurr) % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
}
else
{
eSliceType = (pocLast == 0 || (pocCurr - (isField ? 1 : 0)) % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
}
}
rpcSlice->setSliceType ( eSliceType );
}
if (m_pcCfg->getUseRecalculateQPAccordingToLambda())
{
dQP = xGetQPValueAccordingToLambda( dLambda );
iQP = max( -rpcSlice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA), min( MAX_QP, (int) floor( dQP + 0.5 ) ) );
}
rpcSlice->setSliceQp ( iQP );
rpcSlice->setSliceQpDelta ( 0 );
#if !W0038_CQP_ADJ
rpcSlice->setSliceChromaQpDelta( COMPONENT_Cb, 0 );
rpcSlice->setSliceChromaQpDelta( COMPONENT_Cr, 0 );
#endif
rpcSlice->setUseChromaQpAdj( rpcSlice->getPPS()->getPpsRangeExtension().getChromaQpOffsetListEnabledFlag() );
rpcSlice->setNumRefIdx(REF_PIC_LIST_0,m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive);
rpcSlice->setNumRefIdx(REF_PIC_LIST_1,m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive);
if ( m_pcCfg->getDeblockingFilterMetric() )
{
rpcSlice->setDeblockingFilterOverrideFlag(true);
rpcSlice->setDeblockingFilterDisable(false);
rpcSlice->setDeblockingFilterBetaOffsetDiv2( 0 );
rpcSlice->setDeblockingFilterTcOffsetDiv2( 0 );
}
else if (rpcSlice->getPPS()->getDeblockingFilterControlPresentFlag())
{
rpcSlice->setDeblockingFilterOverrideFlag( rpcSlice->getPPS()->getDeblockingFilterOverrideEnabledFlag() );
rpcSlice->setDeblockingFilterDisable( rpcSlice->getPPS()->getPPSDeblockingFilterDisabledFlag() );
if ( !rpcSlice->getDeblockingFilterDisable())
{
if ( rpcSlice->getDeblockingFilterOverrideFlag() && eSliceType!=I_SLICE)
{
rpcSlice->setDeblockingFilterBetaOffsetDiv2( m_pcCfg->getGOPEntry(iGOPid).m_betaOffsetDiv2 + m_pcCfg->getLoopFilterBetaOffset() );
rpcSlice->setDeblockingFilterTcOffsetDiv2( m_pcCfg->getGOPEntry(iGOPid).m_tcOffsetDiv2 + m_pcCfg->getLoopFilterTcOffset() );
}
else
{
rpcSlice->setDeblockingFilterBetaOffsetDiv2( m_pcCfg->getLoopFilterBetaOffset() );
rpcSlice->setDeblockingFilterTcOffsetDiv2( m_pcCfg->getLoopFilterTcOffset() );
}
}
}
else
{
rpcSlice->setDeblockingFilterOverrideFlag( false );
rpcSlice->setDeblockingFilterDisable( false );
rpcSlice->setDeblockingFilterBetaOffsetDiv2( 0 );
rpcSlice->setDeblockingFilterTcOffsetDiv2( 0 );
}
rpcSlice->setDepth ( depth );
pcPic->layer = temporalId;
if(eSliceType==I_SLICE)
{
pcPic->layer = 0;
}
rpcSlice->setTLayer( pcPic->layer );
rpcSlice->setSliceMode ( m_pcCfg->getSliceMode() );
rpcSlice->setSliceArgument ( m_pcCfg->getSliceArgument() );
#if HEVC_DEPENDENT_SLICES
rpcSlice->setSliceSegmentMode ( m_pcCfg->getSliceSegmentMode() );
rpcSlice->setSliceSegmentArgument ( m_pcCfg->getSliceSegmentArgument() );
#endif
rpcSlice->setMaxNumMergeCand ( m_pcCfg->getMaxNumMergeCand() );
rpcSlice->setMaxBTSize ( rpcSlice->isIntra() ? MAX_BT_SIZE : MAX_BT_SIZE_INTER );
}
#if SHARP_LUMA_DELTA_QP
double EncSlice::calculateLambda( const Slice* slice,
const int GOPid, // entry in the GOP table
const int depth, // slice GOP hierarchical depth.
const double refQP, // initial slice-level QP
const double dQP, // initial double-precision QP
int &iQP ) // returned integer QP.
{
enum SliceType eSliceType = slice->getSliceType();
const bool isField = slice->getPic()->fieldPic;
const int NumberBFrames = ( m_pcCfg->getGOPSize() - 1 );
const int SHIFT_QP = 12;
#if X0038_LAMBDA_FROM_QP_CAPABILITY
const int temporalId=m_pcCfg->getGOPEntry(GOPid).m_temporalId;
const std::vector<double> &intraLambdaModifiers=m_pcCfg->getIntraLambdaModifier();
#endif
#if DISTORTION_LAMBDA_BUGFIX
int bitdepth_luma_qp_scale = 6
* (slice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8
- DISTORTION_PRECISION_ADJUSTMENT(slice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA)));
#else
#if FULL_NBIT
int bitdepth_luma_qp_scale = 6 * (slice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8);
#else
int bitdepth_luma_qp_scale = 0;
#endif
#endif
double qp_temp = dQP + bitdepth_luma_qp_scale - SHIFT_QP;
// Case #1: I or P-slices (key-frame)
double dQPFactor = m_pcCfg->getGOPEntry(GOPid).m_QPFactor;
if ( eSliceType==I_SLICE )
{
if (m_pcCfg->getIntraQpFactor()>=0.0 && m_pcCfg->getGOPEntry(GOPid).m_sliceType != I_SLICE)
{
dQPFactor=m_pcCfg->getIntraQpFactor();
}
else
{
#if X0038_LAMBDA_FROM_QP_CAPABILITY
if(m_pcCfg->getLambdaFromQPEnable())
{
dQPFactor=0.57;
}
else
{
#endif
double dLambda_scale = 1.0 - Clip3( 0.0, 0.5, 0.05*(double)(isField ? NumberBFrames/2 : NumberBFrames) );
dQPFactor=0.57*dLambda_scale;
#if X0038_LAMBDA_FROM_QP_CAPABILITY
}
#endif
}
}
#if X0038_LAMBDA_FROM_QP_CAPABILITY
else if( m_pcCfg->getLambdaFromQPEnable() )
{
dQPFactor=0.57;
}
#endif
double dLambda = dQPFactor*pow( 2.0, qp_temp/3.0 );
#if X0038_LAMBDA_FROM_QP_CAPABILITY
if( !(m_pcCfg->getLambdaFromQPEnable()) && depth>0 )
#else
if ( depth>0 )
#endif
{
#if DISTORTION_LAMBDA_BUGFIX
double qp_temp_ref = refQP + bitdepth_luma_qp_scale - SHIFT_QP;
dLambda *= Clip3(2.00, 4.00, (qp_temp_ref / 6.0)); // (j == B_SLICE && p_cur_frm->layer != 0 )
#else
#if FULL_NBIT
double qp_temp_ref_orig = refQP - SHIFT_QP;
dLambda *= Clip3( 2.00, 4.00, (qp_temp_ref_orig / 6.0) ); // (j == B_SLICE && p_cur_frm->layer != 0 )
#else
double qp_temp_ref = refQP + bitdepth_luma_qp_scale - SHIFT_QP;
dLambda *= Clip3( 2.00, 4.00, (qp_temp_ref / 6.0) ); // (j == B_SLICE && p_cur_frm->layer != 0 )
#endif
#endif
}
// if hadamard is used in ME process
if ( !m_pcCfg->getUseHADME() && slice->getSliceType( ) != I_SLICE )
{
dLambda *= 0.95;
}
#if X0038_LAMBDA_FROM_QP_CAPABILITY
double lambdaModifier;
if( eSliceType != I_SLICE || intraLambdaModifiers.empty())
{
lambdaModifier = m_pcCfg->getLambdaModifier( temporalId );
}
else
{
lambdaModifier = intraLambdaModifiers[ (temporalId < intraLambdaModifiers.size()) ? temporalId : (intraLambdaModifiers.size()-1) ];
}
dLambda *= lambdaModifier;
#endif
iQP = max( -slice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA), min( MAX_QP, (int) floor( dQP + 0.5 ) ) );
#if JVET_K0072
if( m_pcCfg->getDepQuantEnabledFlag() )
{
dLambda *= pow( 2.0, 0.25/3.0 ); // slight lambda adjustment for dependent quantization (due to different slope of quantizer)
}
#endif
// NOTE: the lambda modifiers that are sometimes applied later might be best always applied in here.
return dLambda;
}
#endif
void EncSlice::resetQP( Picture* pic, int sliceQP, double lambda )
{
Slice* slice = pic->slices[0];
// store lambda
slice->setSliceQp( sliceQP );
setUpLambda(slice, lambda, sliceQP);
}
#if ENABLE_QPA
static inline int apprI2Log2 (const double d)
{
return d < 6.0e-20 ? -128 : int(floor(2.0 * log(d) / log(2.0) + 0.5));
}
#ifndef HLM_L1_NORM
#define HLM_L1_NORM
#endif
static int filterAndCalculateAverageEnergies (const Pel* pSrc, const int iSrcStride,
double &hpEner, const int iHeight, const int iWidth,
const int iPOC = 0)
{
int iHpValue;
uint32_t uHpERow, uHpEner = 0;
// skip first row as there may be a black border frame
pSrc += iSrcStride;
// center rows
for (int y = 1; y < iHeight - 1; y++)
{
uHpERow = 0;
// skip column as there may be a black border frame
for (int x = 1; x < iWidth - 1; x++) // and columns
{
iHpValue = 4 * (int)pSrc[x] - (int)pSrc[x-1] - (int)pSrc[x+1] - (int)pSrc[x-iSrcStride] - (int)pSrc[x+iSrcStride];
#ifdef HLM_L1_NORM
uHpERow += abs (iHpValue);
#else
uHpERow += iHpValue * iHpValue;
#endif
}
// skip column as there may be a black border frame
#ifdef HLM_L1_NORM
uHpEner += uHpERow;
#else
uHpEner += (uHpERow + 64) >> 7; // avoids overflows
#endif
pSrc += iSrcStride;
}
// skip last row as there may be a black border frame
hpEner = double(uHpEner) / double((iWidth - 2) * (iHeight - 2));
#ifdef HLM_L1_NORM
hpEner *= hpEner;
#endif
// lower limit, compensate for highpass amplification
if (hpEner < 64.0) hpEner = 64.0;
if (iPOC <= 0) return 0;
return 1; // OK
}
#ifdef HLM_L1_NORM
#undef HLM_L1_NORM
#endif
#if ENABLE_QPA
static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, const PreCalcValues& pcv,
const uint32_t startAddr, const uint32_t boundingAddr, const bool useSharpLumaDQP,
const int gopSize, const double hpEnerAvg, const double hpEnerMax)
{
const int iBitDepth = pcSlice->getSPS()->getBitDepth (CHANNEL_TYPE_LUMA);
const int iQPIndex = pcSlice->getSliceQp(); // initial QP index for current slice, used in following loops
#if HEVC_TILES_WPP
const TileMap& tileMap = *pcPic->tileMap;
#endif
bool sliceQPModified = false;
double hpEnerPic = 1.0 / (1.5 * double(1 << iBitDepth)); // speedup: multiply instead of divide in loops below
if (pcv.lumaWidth > 2048 && pcv.lumaHeight > 1280) // for UHD/4K
{
hpEnerPic *= 1.5;
}
if ((pcPic->getPOC() & 1) && (iQPIndex >= MAX_QP))
{
int iQPFixed = Clip3 (0, MAX_QP, iQPIndex + ((apprI2Log2 (hpEnerAvg * hpEnerPic) + apprI2Log2 (hpEnerMax * hpEnerPic) + 1) >> 1)); // adapted slice QP = (mean(QP) + max(QP)) / 2
#if SHARP_LUMA_DELTA_QP
// change new fixed QP based on average CTU luma value (Sharp)
if (useSharpLumaDQP)
{
uint64_t uAvgLuma = 0;
for (uint32_t ctuTsAddr = startAddr; ctuTsAddr < boundingAddr; ctuTsAddr++)
{
#if HEVC_TILES_WPP
const uint32_t ctuRsAddr = tileMap.getCtuTsToRsAddrMap (ctuTsAddr);
#else
const uint32_t ctuRsAddr = ctuTsAddr;
#endif
uAvgLuma += (uint64_t)pcPic->m_iOffsetCtu[ctuRsAddr];
}
uAvgLuma = (uAvgLuma + ((boundingAddr - startAddr) >> 1)) / (boundingAddr - startAddr);
iQPFixed = Clip3 (0, MAX_QP, iQPFixed + 1 - int((3 * uAvgLuma * uAvgLuma) >> uint64_t(2 * iBitDepth - 1)));
}
#endif
if (iQPFixed < iQPIndex) iQPFixed = iQPIndex;
else
if (iQPFixed > iQPIndex)
{
const double* oldLambdas = pcSlice->getLambdas();
const double corrFactor = pow (2.0, double(iQPFixed - iQPIndex) / 3.0);
const double newLambdas[MAX_NUM_COMPONENT] = {oldLambdas[0] * corrFactor, oldLambdas[1] * corrFactor, oldLambdas[2] * corrFactor};
CHECK (iQPIndex != pcSlice->getSliceQpBase(), "Invalid slice QP!");
pcSlice->setLambdas (newLambdas);
pcSlice->setSliceQp (iQPFixed); // update the slice/base QPs
pcSlice->setSliceQpBase (iQPFixed);
sliceQPModified = true;
}
for (uint32_t ctuTsAddr = startAddr; ctuTsAddr < boundingAddr; ctuTsAddr++)
{
#if HEVC_TILES_WPP
const uint32_t ctuRsAddr = tileMap.getCtuTsToRsAddrMap (ctuTsAddr);
#else
const uint32_t ctuRsAddr = ctuTsAddr;
#endif
pcPic->m_iOffsetCtu[ctuRsAddr] = (Pel)iQPFixed; // fixed QPs
}
}
else
{
for (uint32_t ctuTsAddr = startAddr; ctuTsAddr < boundingAddr; ctuTsAddr++)
{
#if HEVC_TILES_WPP
const uint32_t ctuRsAddr = tileMap.getCtuTsToRsAddrMap (ctuTsAddr);
#else
const uint32_t ctuRsAddr = ctuTsAddr;
#endif
int iQPAdapt = Clip3 (0, MAX_QP, iQPIndex + apprI2Log2 (pcPic->m_uEnerHpCtu[ctuRsAddr] * hpEnerPic));
#if SHARP_LUMA_DELTA_QP
if ((pcv.widthInCtus > 1) && (gopSize > 1)) // try to enforce CTU SNR greater than zero dB
#else
if ((!pcSlice->isIntra()) && (gopSize > 1)) // try to enforce CTU SNR greater than zero dB
#endif
{
const Pel dcOffset = pcPic->m_iOffsetCtu[ctuRsAddr];
#if SHARP_LUMA_DELTA_QP
// change adaptive QP based on mean CTU luma value (Sharp)
if (useSharpLumaDQP)
{
const uint64_t uAvgLuma = (uint64_t)dcOffset;
iQPAdapt = max (0, iQPAdapt + 1 - int((3 * uAvgLuma * uAvgLuma) >> uint64_t(2 * iBitDepth - 1)));
}
#endif
const uint32_t uRefScale = g_invQuantScales[iQPAdapt % 6] << ((iQPAdapt / 6) + iBitDepth - (pcSlice->isIntra() ? 4 : 3));
const CompArea subArea = clipArea (CompArea (COMPONENT_Y, pcPic->chromaFormat, Area ((ctuRsAddr % pcv.widthInCtus) * pcv.maxCUWidth, (ctuRsAddr / pcv.widthInCtus) * pcv.maxCUHeight, pcv.maxCUWidth, pcv.maxCUHeight)), pcPic->Y());
const Pel* pSrc = pcPic->getOrigBuf (subArea).buf;
const SizeType iSrcStride = pcPic->getOrigBuf (subArea).stride;
const SizeType iSrcHeight = pcPic->getOrigBuf (subArea).height;
const SizeType iSrcWidth = pcPic->getOrigBuf (subArea).width;
uint32_t uAbsDCless = 0;
// compute sum of absolute DC-less (high-pass) luma values
for (SizeType h = 0; h < iSrcHeight; h++)
{
for (SizeType w = 0; w < iSrcWidth; w++)
{
uAbsDCless += (uint32_t)abs (pSrc[w] - dcOffset);
}
pSrc += iSrcStride;
}
if (iSrcHeight >= 64 || iSrcWidth >= 64) // normalization
{
const uint64_t blockSize = uint64_t(iSrcWidth * iSrcHeight);
uAbsDCless = uint32_t((uint64_t(uAbsDCless) * 64*64 + (blockSize >> 1)) / blockSize);
}
if (uAbsDCless < 64*64) uAbsDCless = 64*64; // limit to 1
// reduce QP index if CTU would be fully quantized to zero
if (uAbsDCless < uRefScale)
{
const int limit = min (0, ((iQPIndex + 4) >> 3) - 6);
const int redVal = max (limit, apprI2Log2 ((double)uAbsDCless / (double)uRefScale));
iQPAdapt = max (0, iQPAdapt + redVal);
}
#if SHARP_LUMA_DELTA_QP
if (iQPAdapt > MAX_QP) iQPAdapt = MAX_QP;
#endif
}
pcPic->m_iOffsetCtu[ctuRsAddr] = (Pel)iQPAdapt; // adapted QPs
if ((pcv.widthInCtus > 1) && (gopSize > 1)) // try to reduce local bitrate peaks via minimum smoothing
{
iQPAdapt = ctuRsAddr % pcv.widthInCtus; // horizontal offset
if (iQPAdapt == 0)
{
iQPAdapt = (ctuRsAddr > 1) ? pcPic->m_iOffsetCtu[ctuRsAddr - 2] : 0;
}
else // iQPAdapt >= 1
{
iQPAdapt = (iQPAdapt > 1) ? min (pcPic->m_iOffsetCtu[ctuRsAddr - 2], pcPic->m_iOffsetCtu[ctuRsAddr]) : pcPic->m_iOffsetCtu[ctuRsAddr];
}
if (ctuRsAddr > pcv.widthInCtus)
{
iQPAdapt = min (iQPAdapt, (int)pcPic->m_iOffsetCtu[ctuRsAddr - 1 - pcv.widthInCtus]); // min(L, T)
}
if ((ctuRsAddr > 0) && (pcPic->m_iOffsetCtu[ctuRsAddr - 1] < (Pel)iQPAdapt))
{
pcPic->m_iOffsetCtu[ctuRsAddr - 1] = (Pel)iQPAdapt;
}
if ((ctuTsAddr == boundingAddr - 1) && (ctuRsAddr > pcv.widthInCtus)) // last CTU in the given slice
{
iQPAdapt = min (pcPic->m_iOffsetCtu[ctuRsAddr - 1], pcPic->m_iOffsetCtu[ctuRsAddr - pcv.widthInCtus]);
if (pcPic->m_iOffsetCtu[ctuRsAddr] < (Pel)iQPAdapt)
{
pcPic->m_iOffsetCtu[ctuRsAddr] = (Pel)iQPAdapt;
}
}
}
} // end iteration over all CTUs in current slice
}
return sliceQPModified;
}
#endif // ENABLE_QPA
#endif // ENABLE_QPA || ENABLE_PRIVATE
// ====================================================================================================================
// Public member functions
// ====================================================================================================================
//! set adaptive search range based on poc difference
void EncSlice::setSearchRange( Slice* pcSlice )
{
int iCurrPOC = pcSlice->getPOC();
int iRefPOC;
int iGOPSize = m_pcCfg->getGOPSize();
int iOffset = (iGOPSize >> 1);
int iMaxSR = m_pcCfg->getSearchRange();
int iNumPredDir = pcSlice->isInterP() ? 1 : 2;
for (int iDir = 0; iDir < iNumPredDir; iDir++)
{
RefPicList e = ( iDir ? REF_PIC_LIST_1 : REF_PIC_LIST_0 );
for (int iRefIdx = 0; iRefIdx < pcSlice->getNumRefIdx(e); iRefIdx++)
{
iRefPOC = pcSlice->getRefPic(e, iRefIdx)->getPOC();
int newSearchRange = Clip3(m_pcCfg->getMinSearchWindow(), iMaxSR, (iMaxSR*ADAPT_SR_SCALE*abs(iCurrPOC - iRefPOC)+iOffset)/iGOPSize);
m_pcInterSearch->setAdaptiveSearchRange(iDir, iRefIdx, newSearchRange);
#if ENABLE_WPP_PARALLELISM
for( int jId = 1; jId < m_pcLib->getNumCuEncStacks(); jId++ )
{
m_pcLib->getInterSearch( jId )->setAdaptiveSearchRange( iDir, iRefIdx, newSearchRange );
}
#endif
}
}
}
/**
Multi-loop slice encoding for different slice QP
\param pcPic picture class
*/
void EncSlice::precompressSlice( Picture* pcPic )
{
// if deltaQP RD is not used, simply return
if ( m_pcCfg->getDeltaQpRD() == 0 )
{
return;
}
if ( m_pcCfg->getUseRateCtrl() )
{
THROW("\nMultiple QP optimization is not allowed when rate control is enabled." );
}
Slice* pcSlice = pcPic->slices[getSliceSegmentIdx()];
#if HEVC_DEPENDENT_SLICES
if (pcSlice->getDependentSliceSegmentFlag())
{
// if this is a dependent slice segment, then it was optimised
// when analysing the entire slice.
return;
}
#endif
if (pcSlice->getSliceMode()==FIXED_NUMBER_OF_BYTES)
{
// TODO: investigate use of average cost per CTU so that this Slice Mode can be used.
THROW( "Unable to optimise Slice-level QP if Slice Mode is set to FIXED_NUMBER_OF_BYTES\n" );
}
double dPicRdCostBest = MAX_DOUBLE;
uint32_t uiQpIdxBest = 0;
double dFrameLambda;
#if DISTORTION_LAMBDA_BUGFIX
int SHIFT_QP = 12
+ 6
* (pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8
- DISTORTION_PRECISION_ADJUSTMENT(pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA)));
#else
#if FULL_NBIT
int SHIFT_QP = 12 + 6 * (pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8);
#else
int SHIFT_QP = 12;
#endif