![Versatile Video Coding | Panel on Emerging Video Codecs | Picture Coding Symposium 2019 | Ningbo, China
15.11.2019 | Mathias Wien | Institute of Imaging and Computer Vision | RWTH Aachen University
5
• Assessment according to JVET Common Testing Conditions [JVET-M1010]
Rate-distortion performance, encoder / decoder run-time of VVC reference software implementation
• Systematic evaluation throughout standardization project
All Intra, Random Access, Low Delay B configurations
Comparison to HEVC reference software HM: Comparable rate-distortion based encoder implementations
Versatile Video Coding – Compression Performance](https://image.slidesharecdn.com/pcs2019panelemergingvideocodingstandardsvvcmwi-191126134131/85/PCS-2019-Panel-on-Emerging-Video-Coding-Standards-Overview-on-the-Emerging-Versatile-Video-Coding-Standard-5-320.jpg)
PCS 2019 Panel on Emerging Video Coding Standards: Overview on the Emerging Versatile Video Coding Standard
- 1. Brief Overview on the Emerging Versatile Video Coding Standard … as seen by an individual Picture Coding Symposium 2019, Ningbo, China Mathias Wien Lehrstuhl für Bildverarbeitung RWTH Aachen University [email protected]
- 2. Versatile Video Coding | Panel on Emerging Video Codecs | Picture Coding Symposium 2019 | Ningbo, China 15.11.2019 | Mathias Wien | Institute of Imaging and Computer Vision | RWTH Aachen University 2 • Joint Video Experts Team of ISO/IEC MPEG and ITU-T VCEG, est. in Oct. 2015 as Joint Video Exploration Team, renamed Apr. 2018 at CfP eval. • Joint Exploration Model: 7 iterations Investigating coding tools enhanced relative to HEVC test model HM • Call for Evidence (CfE), issued Apr. 2017, evaluated in Jul. 2017 Significant improvement for JEM and independent submissions • Call for Proposls (CfP), issued Oct. 2017, evaluated in Apr. 2018 at10th JVET meeting Kick-off for VVC, definition of name, VVC draft 1 and test model 1 (VTM1) Evolution towards VVC Evolution of JVET docs and participants
- 3. Versatile Video Coding | Panel on Emerging Video Codecs | Picture Coding Symposium 2019 | Ningbo, China 15.11.2019 | Mathias Wien | Institute of Imaging and Computer Vision | RWTH Aachen University 3 • Entropy Coding CABAC Multi-hypothesis probability estimation • Coding structures 4 × 4 to 128 × 128 Coding Unit (CU) size Quad+ternary+binary tree partitioning Dual-tree (separate trees for luma/chroma) Partial coverage of prediction regions by transforms (SBT,ISP) • Inter Coding 16th-sample prec. for derived MVs | Locally adaptive MV resolution: signaling at ¼, 1, 4 sample precision 8-tap DCTIF + alt. 6-tap switched IF for ½ sample, 6-tap for affine subblock MC Block-level CU weights, Triangular prediction with merge Combined inter/intra prediction (CIIP) Affine prediction refinement with optical flow (PROF), Bi-directional optical flow prediction (BDOF) Subblock-based temporal merge, affine MV control point MVP, decoder-side MV refinement, history-based MVP, symmetrical MVD, Merge mode with MVD (MMVD) • Intra Coding DC, planar, 67 intra prediction directions signaled, 93 after wide angle mapping | signaling: 6 most probable modes + remainder Matrix-based intra prediction Multi reference lines intra prediction Intra subpartitions (ISP) Position-dependent intra prediction sample filtering Cross-component linear prediction • Residual Coding DCT2 4×4-64×64, multi-transform selection (DST7/DCT8, ≤32), inter: subblock transform (SBT), low-frequency non-separable transform (LFNST), transform skip, 4:4:4 adaptive colour transform Coefficient coding: HEVC + parity-based binarization for dependent quantization (DQ) • Loop Filters Deblocking filter, sample adaptive offset (SAO) Adaptive loop filter (ALF), luma mapping with chroma scaling VVC Coding Tools in a Nutshell MV/MVP/MVD: motion vector / predictor / difference | DCTIF: Discrete cosine transform derived interpolation filter | DST: Discrete sine transform
- 4. Versatile Video Coding | Panel on Emerging Video Codecs | Picture Coding Symposium 2019 | Ningbo, China 15.11.2019 | Mathias Wien | Institute of Imaging and Computer Vision | RWTH Aachen University 4 • Parameter sets Video parameter set, sequence parameter set, picture parameter set (VPS, SPS, PPS) Adaptation parameter set (APS), picture header • Picture types IDR, CRA, RADL, RASL, Trail, STSA Gradual decoding refresh (GDR) • Picture partitioning Slices, tiles Subpictures, raster-scan slices of multiple tiles, rectangular slices within tiles • Screen content coding Palette mode for 4:4:4 Intra block copy (IBC) Residual coding for transform skip Block-based Delta PCM (BDPCM) • More versatility Reference picture resampling (RPR) Spatial, quality and temporal scalability Bitstream extraction and merging without VCL NAL unit rewriting Film grain synthesis SEI message VVC Versatility Aspects IDR: instantaneous decoding refresh | CRA: clean random access | RADL/RASL: random access decodable/skipped leading pic. | STSA: step-wise temporal sublayer access
- 5. Versatile Video Coding | Panel on Emerging Video Codecs | Picture Coding Symposium 2019 | Ningbo, China 15.11.2019 | Mathias Wien | Institute of Imaging and Computer Vision | RWTH Aachen University 5 • Assessment according to JVET Common Testing Conditions [JVET-M1010] Rate-distortion performance, encoder / decoder run-time of VVC reference software implementation • Systematic evaluation throughout standardization project All Intra, Random Access, Low Delay B configurations Comparison to HEVC reference software HM: Comparable rate-distortion based encoder implementations Versatile Video Coding – Compression Performance
- 6. Thanks for your attention! Mathias Wien Lehrstuhl für Bildverarbeitung RWTH Aachen University [email protected]