Decoder Side Motion Vector Derivation for Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in efficiently compressing high-resolution videos due to increased bandwidth demands, with existing standards like HEVC struggling to optimize motion vector derivation and prediction modes for improved coding efficiency.
Innovation Solution
The implementation of decoder-side motion vector derivation (DMVD) tools that selectively enable refinement of motion information in bitstream representations, applying advanced prediction modes such as multi-hypothesis, asymmetric weighting, and bi-directional affine modes to enhance video processing and compression performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If decoder-side motion vector derivation (DMVD) tool is enabled for all blocks, then motion information refinement improves coding efficiency, but computational complexity increases
Solution Approach 1:
The patent applies local quality by selectively enabling DMVD tool only for specific block types (bi-prediction blocks, AMVP blocks, affine blocks) rather than uniformly applying it to all blocks. This allows motion information refinement to be focused where it provides the most benefit while avoiding unnecessary computational overhead in blocks where it would be less effective.
Solution Approach 2:
The patent segments the video blocks into different categories based on their prediction mode (bi-prediction, AMVP, affine, etc.) and applies DMVD selectively to each segment. This segmentation approach allows the system to optimize the balance between coding efficiency improvement and computational complexity by treating different block types differently.
2Measurement precision
If multi-hypothesis prediction mode is applied to all blocks, then prediction accuracy improves, but processing time and complexity increase
Solution Approach 1:
The patent applies multi-hypothesis prediction mode selectively to specific block types (bi-prediction blocks, AMVP blocks, affine blocks) rather than uniformly to all blocks. This localized application improves prediction accuracy for blocks that benefit most from multiple motion hypotheses while avoiding the processing overhead for blocks where simple prediction suffices.
Solution Approach 2:
The patent dynamically selects whether to apply multi-hypothesis prediction based on the block's characteristics and prediction mode. The system adapts its processing approach for each block type, applying complex multi-hypothesis analysis only where needed and using simpler methods elsewhere, thereby optimizing the trade-off between prediction accuracy and processing time.
3Measurement precision
If asymmetric weighting factors are applied to all reference blocks, then prediction precision improves, but computational overhead increases
Solution Approach 1:
The patent applies asymmetric weighting factors selectively to specific reference blocks based on their reliability and quality metrics. Rather than uniformly applying asymmetric weighting to all reference blocks, the system identifies which reference blocks benefit most from asymmetric treatment and applies the computational overhead only in those cases, improving prediction precision efficiently.
4Measurement precision
If sub-block level DMVD is enabled, then motion refinement precision improves, but processing complexity increases
Solution Approach 1:
The patent segments the current block into sub-blocks and applies DMVD at the sub-block level selectively based on block characteristics. This segmentation allows the system to achieve higher motion refinement precision in regions with complex motion patterns while avoiding the processing complexity in regions with simpler motion, thereby optimizing the trade-off between precision and complexity.
Data Source
AI summary
Devices, systems and methods for digital video coding, which include decoder side motion vector derivation (DMVD) tools, are described. An exemplary method for video processing includes making a decision, based on a determination that a current block of a video is coded using a multi-hypothesis prediction mode, regarding a selective enablement of a DMVD tool for the current block, wherein the DMVD tool derives a refinement of motion information signaled in a bitstream representation of the video; and performing, based on the decision, a conversion between the current block and the bitstream representation.