Merge Candidate Reordering for Lower-Bit MMVD Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards like HEVC face inefficiencies in motion vector coding, particularly in Merge Mode, due to redundant candidate prediction positions and high bit usage for signaling candidate indices.
Innovation Solution
A video coder reorders candidate prediction positions based on computed costs, reducing the number of bits needed for coding by selecting a subset of the lowest cost positions, and applies this method to Merge Mode with Motion Vector Difference (MMVD) in VVC, refining motion information with motion vector differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all candidate prediction positions are coded in HEVC Merge mode, then prediction accuracy is maintained, but bit usage for signaling candidate indices increases
Solution Approach 1:
The patent extracts and removes redundant candidate prediction positions from the candidate list. By identifying and eliminating positions that provide minimal or duplicate prediction value, the system reduces the number of candidates that need to be signaled, thereby reducing bit usage while maintaining prediction accuracy for the essential candidates.
Solution Approach 2:
The patent segments the candidate prediction positions into different groups or categories based on their prediction characteristics. This segmentation allows the system to prioritize and efficiently encode the most important candidate groups while using fewer bits for less critical segments, resolving the contradiction between maintaining accuracy and reducing bit usage.
2Adaptability or versatility
If redundant candidate prediction positions are included in Merge mode, then prediction coverage is improved, but coding efficiency decreases
Solution Approach 1:
The patent applies partial action by including only the necessary subset of candidate prediction positions rather than all possible candidates. By carefully selecting and ordering candidates to include just enough coverage for most prediction scenarios, the system achieves adequate prediction coverage while significantly improving coding efficiency by reducing the number of candidates that require encoding.
Solution Approach 2:
The patent changes the parameter of candidate list size and composition dynamically. By adjusting which candidates are included and their ordering based on prediction costs and characteristics, the system optimizes the balance between prediction coverage and coding efficiency, achieving high versatility with reduced complexity.
3Measurement precision
If candidate indices are signaled with high precision in bitstream, then candidate selection accuracy is improved, but bitrate increases
Solution Approach 1:
The patent performs preliminary ordering and selection of candidate prediction positions before encoding. By pre-organizing candidates in an optimal sequence and identifying the most important ones in advance, the system can signal candidate indices with fewer bits while maintaining accurate selection, as the preliminary arrangement reduces the entropy of the index values that need to be encoded.
Data Source
AI summary
A video coding system that reorders prediction candidates is provided. A video coder receives data for a block of pixels to be encoded or decoded as a current block of a current picture of a video. The video coder identifies possible candidate prediction positions. The video coder computes a cost for each of the identified possible candidate prediction positions. The video coder assigns, based on the computed costs, a reordered index to each of N lowest cost candidate prediction positions from the identified possible candidate prediction positions. The video coder selects a candidate prediction position using the assigned reordered indices, wherein the selection is signaled in or parsed from the bitstream. The video coder encodes or decodes the current block by using the selected candidate prediction position.