DMVR Control for Reference Picture Resampling in Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies when dealing with differences in spatial resolution between current and reference pictures, leading to unnecessary resource usage and reduced user experience.
Innovation Solution
Implementing decoder-side motion vector refinement (DMVR) that can be selectively enabled or disabled based on the spatial resolution match between current and reference pictures, allowing for improved coding efficiency by reducing processor, memory, and network resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If DMVR is enabled for all blocks when reference picture resampling is used, then motion vector refinement is improved, but processor and memory resource usage increases unnecessarily
Solution Approach 1:
The patent applies local quality by enabling DMVR selectively only for specific block types (intra-block copy blocks and inter-blocks with specific motion vector differences) rather than uniformly for all blocks. This localized application of the refinement process ensures that computational resources are concentrated where they provide the most benefit, avoiding unnecessary processing for blocks where DMVR would not improve accuracy.
Solution Approach 2:
The patent implements partial action by applying DMVR to only a subset of blocks that meet specific criteria (block type and motion vector difference thresholds) rather than performing the operation on all blocks. This partial application strategy achieves sufficient motion vector refinement for critical blocks while avoiding the excessive resource consumption that would result from universal application.
2Use of energy by moving object
If DMVR is disabled when spatial resolutions differ, then resource usage is reduced, but coding efficiency deteriorates
Solution Approach 1:
The patent applies local quality by making DMVR application block-specific rather than picture-resolution-dependent. Each block is evaluated individually based on its type and motion characteristics, allowing DMVR to be enabled for specific blocks even when the overall picture resolution differs from reference pictures. This selective approach maintains coding efficiency for critical blocks while avoiding universal disabling.
Solution Approach 2:
The patent implements dynamics by making the DMVR enablement decision adaptive to each block's specific characteristics (block type, motion vector difference) rather than using a static decision based solely on picture resolution. This dynamic, context-dependent approach allows the system to optimize coding efficiency on a per-block basis while managing overall resource consumption.
3Productivity
If DMVR is selectively enabled based on block type and motion vector difference, then coding efficiency is improved, but device complexity increases
Solution Approach 1:
The patent applies parameter changes by using simple, easily obtainable block attributes (block type classification and motion vector difference values) as decision criteria for enabling DMVR. These parameters are naturally available during the decoding process and require minimal additional computation to evaluate, allowing selective DMVR enablement without significantly increasing device complexity.
Solution Approach 2:
The patent applies local quality by implementing straightforward decision logic at each block level that checks basic block properties and motion characteristics. This localized decision-making process uses simple conditional checks rather than complex global analysis, enabling improved coding efficiency through selective DMVR while keeping the added complexity minimal and localized to individual block processing.
Data Source
AI summary
A method of decoding includes determining, by the video decoder, whether a resolution of a current picture being decoded is the same as the resolution of reference pictures identified by a reference picture list associated with the current picture; enabling, by the video decoder, decoder-side motion vector refinement (DMVR) for a current block of the current picture when the resolution of the current picture is determined to be the same as the resolution of each of the reference pictures; disabling, by the video decoder, the DMVR for the current block of the current picture when the resolution of the current picture is determined to be different than the resolution of either of the reference pictures; and refining, by the video decoder, motion vectors corresponding to the current block using the DMVR when the DMVR flag is enabled for the current block.


