Boosted Affine Motion Compensation For Geometric Partitioning Mode
Geometric partitioning mode (GPM) is an inter coding tool developed for versatile video coding (VVC), which is utilized to split a coding block into two sub-partitions geometrically. These two GPM sub-partitions are predicted separately using translational motion compensation and subsequently blended. Recently, GPM with affine prediction (GPM-affine) has been introduced in the enhanced compression model (ECM) to facilitate affine motion compensation for predicting the GPM sub-partitions. To further enhance the efficiency of GPM-affine, a method of boosted affine motion compensation for geometric partitioning mode (BAMC-GPM) is proposed in this paper. Specifically, the affine motion information in GPM-affine can be refined using a template-based approach, where a motion vector offset is derived based on template costs and utilized to update the control point motion vectors. Experimental results demonstrate that BAMC-GPM achieves average Bjontegaard Delta rate (BD-rate) savings of 0.04% and 0.12% compared to ECM-14.0 under random access (RA) and low delay B (LDB) configurations, respectively. It is noted that BAMC-GPM has been adopted into ECM during the 36th Joint Video Experts Team (JVET) meeting in 2024.
Authors
- Yang Wang (ORCID: https://orcid.org/0000-0003-1029-9280)
- Kai Zhang (ORCID: https://orcid.org/0000-0002-6627-0009)
- Zhang Li (ORCID: https://orcid.org/0009-0003-8559-4599)
Publication Details
- Published
- 2025-08-18
- DOI
- https://doi.org/10.1109/icip55913.2025.11084496
- Primary Topic
- Advanced Vision and Imaging
- Type
- article
- Field-Weighted Citation Impact
- 0.00