Decoder-Side Motion Vector Refinement With Early Exit Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video codecs face challenges in achieving accurate motion vector estimation with reduced computational complexity and signaling overhead, particularly in decoder-side refinement processes.
Innovation Solution
A method for obtaining sub-pixel accurate delta motion vectors through iterative refinement operations using cost functions, including early exit criteria and parametric error surface fitting, to determine motion vectors in decoder-side motion vector refinement systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If iterative refinement operations are performed to improve motion vector accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent applies early exit criteria that allow the iterative refinement process to terminate before completing all planned iterations when a satisfactory motion vector accuracy is achieved. This partial action approach maintains measurement precision while reducing the excessive computational complexity that would result from always performing the maximum number of iterations.
Solution Approach 2:
The patent dynamically adjusts refinement parameters including the number of iterations, search range, and cost function thresholds based on the specific characteristics of each block and the convergence behavior observed during refinement. This adaptive parameter adjustment optimizes the balance between motion vector accuracy and computational complexity for different video content types.
2Measurement precision
If more refinement operations are performed, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The early exit mechanism allows the refinement process to perform only the necessary number of iterations to achieve acceptable accuracy, avoiding unnecessary time consumption from excessive refinement operations. This directly addresses the trade-off between measurement precision and loss of time during decoding.
Solution Approach 2:
The refinement process uses periodic evaluation of cost function improvements to determine whether continued refinement is worthwhile. By checking for convergence at regular intervals and exiting when improvements fall below a threshold, the system maintains measurement precision while minimizing the time lost to overly prolonged refinement operations.
3Measurement precision
If comprehensive search is performed to improve motion vector accuracy, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent segments the search space into multiple refinement stages with progressively smaller search ranges. This segmentation allows comprehensive search to be performed efficiently by dividing it into manageable phases: initial coarse search, intermediate refinement, and final precision refinement. This maintains measurement precision while improving productivity compared to a single comprehensive search.
Solution Approach 2:
The patent performs partial comprehensive search by adapting the search scope based on block characteristics and convergence behavior. When high accuracy is not critical or when convergence is achieved early, the search is truncated to avoid excessive computational work, thereby maintaining productivity while still achieving sufficient measurement precision for the given video content.
Data Source
AI summary
Given that decoder side motion vector refinement/derivation is a normative aspect of a coding system, the encoder will also have to perform the same error surface technique in order to not have any drift between the encoder's reconstruction and the decoder's reconstruction. Hence, all aspects of all embodiments are applicable to both encoding and decoding systems. In template matching, the refinement movement occurs only in the reference starting from the sub-pixel accurate center that is derived based on the explicitly signaled merge index or implicitly through cost evaluations. In bilateral matching (with or without averaged template), the refinements start in the reference lists L0 and L1 starting from the respective sub-pixel accurate centers that are derived based on the explicitly signaled merge index or implicitly through cost evaluations.