Decoder-Side Motion Vector Refinement Memory Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies, such as HEVC, face challenges in achieving efficient compression and memory usage, particularly in the process of decoder side motion vector refinement, where padding is required to avoid overwriting reference samples, leading to increased memory allocation.
Innovation Solution
The proposed solution involves using actual samples loaded from reference images for interpolation in the extended regions during decoder side motion vector refinement, eliminating the need for padding and optimizing memory usage by applying a DMVR motion vector offset to the initial motion vector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If padding is applied to avoid overwriting reference samples during decoder side motion vector refinement, then reliability is improved, but memory allocation increases
Solution Approach 1:
The patent applies preliminary action by performing motion vector refinement using the initial motion vector before the final motion vector is determined. This allows the refinement process to complete using reference samples from the initial position, avoiding the need to pad memory to prevent overwriting during subsequent operations
Solution Approach 2:
The patent inverts the conventional approach by refining the motion vector using the initial motion vector as reference, rather than trying to protect the final motion vector position during refinement. This reversal eliminates the need for padding while maintaining reference sample integrity
2Manufacturing precision
If padding is applied during motion vector refinement, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The refinement operation is performed preliminarily using the initial motion vector, which simplifies the processing by avoiding the need for complex padding operations and multiple memory access patterns that would increase device complexity
Data Source
AI summary
A method, computer program, and computer system is provided for video coding. Video data comprising one or more blocks is received. One or more samples are identified from one or more references images corresponding to the received video data. One or more motion vectors associated with the video data are refined based on interpolating the one or more samples by applying a decoder-side motion vector refinement (DMVR) motion vector offset value to a motion vector associated with a current block from among the one or more blocks.