Decoder-Side Motion Vector Refinement Using Candidate Cost Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently encoding and decoding high-definition and ultra-high-definition video data while maintaining image quality, particularly in reducing redundancy and improving coding efficiency.
Innovation Solution
Implementing decoder-side motion vector refinement (DMVR) techniques that involve deriving an initial motion vector, determining cost values, and refining the motion vector based on updated cost values to improve coding efficiency and image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional motion estimation is used, then decoding speed is maintained at acceptable levels, but coding efficiency and image quality deteriorate
Solution Approach 1:
The patent applies preliminary action by performing motion vector refinement at the decoder side using previously reconstructed reference pictures. The decoder generates multiple candidate motion vectors by searching around the initial motion vector in a predefined search range, and selects the best one based on cost function evaluation. This preliminary refinement process improves coding efficiency without requiring complex encoder-side computation.
Solution Approach 2:
The patent uses copying by creating multiple candidate motion vectors that copy the structure of the initial motion vector but with adjusted components. Each candidate motion vector is constructed by modifying the initial motion vector within a search range, and the best candidate is selected based on reconstruction quality. This copying approach enables refinement while maintaining computational efficiency.
2Measurement precision
If motion vector refinement is performed, then image quality improves, but processing time increases
Solution Approach 1:
The patent applies partial action by performing motion vector refinement only when beneficial, using a cost function to evaluate whether the refinement improves coding efficiency. The search range is limited to a predefined area around the initial motion vector, and the process stops when the best candidate is found. This partial refinement approach improves motion vector precision while controlling processing time.
Solution Approach 2:
The patent implements feedback by using the cost function to evaluate and compare candidate motion vectors against the initial motion vector. The decoder receives feedback from the cost function evaluation and adjusts the motion vector selection accordingly, stopping the refinement process when the initial motion vector is determined to be sufficient. This feedback mechanism balances precision improvement with time consumption.
3Productivity
If multiple motion vector candidates are evaluated, then coding efficiency improves, but computational complexity increases
Solution Approach 1:
The patent applies local quality by focusing the motion vector search around the initial motion vector within a localized search range rather than performing a global search. The cost function evaluation is performed locally on candidate vectors in this restricted range, which improves coding efficiency while keeping computational complexity manageable. The search range and evaluation criteria are adapted to local image characteristics.
Data Source
AI summary
A method for video coding is provided. The method includes: deriving an initial motion vector (MV) of a current block; deriving, by a decoder, an initial motion vector (MV) of a current block; determining, by the decoder, cost values for the initial MV and each of a plurality of MV candidates; obtaining, by the decoder, updated cost values by decreasing a cost value for the initial MV or increasing cost values for the MV candidates; and deriving, by the decoder, a refined MV based on the updated cost values.