HMVP Buffer Reset for Tile-Based Motion Vector Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies do not properly reset the history motion vector predictor (HMVP) buffer when necessary, leading to irrelevant information being retained and affecting compression efficiency.
Innovation Solution
The method involves resetting the HMVP buffer when a block is determined to be at the beginning of a row within a grid segmentation of the video picture, ensuring relevant motion vector predictors are maintained and irrelevant data is removed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the history motion vector buffer is continuously updated without resetting, then the buffer contains more historical motion vectors for prediction, but irrelevant information is retained that reduces prediction accuracy
Solution Approach 1:
The patent applies preliminary action by resetting the history motion vector buffer at specific positions (beginning of rows or picture boundaries) before decoding subsequent blocks. This proactive reset prevents irrelevant motion vectors from accumulating in the buffer, ensuring that only relevant historical motion vectors are retained for prediction. The reset operation is performed in advance at predetermined locations to maintain prediction accuracy throughout the decoding process.
2Productivity
If motion vectors are predicted from neighboring blocks, then the compression ratio is improved, but the prediction efficiency decreases when motion patterns vary significantly
Solution Approach 1:
The patent implements dynamics by adaptively managing the history motion vector buffer based on the decoded block's position and motion characteristics. The buffer is dynamically reset at specific locations (beginning of rows, picture boundaries) and selectively updated with motion vectors from decoded blocks. This dynamic approach allows the system to maintain high compression ratios through motion vector prediction while adapting to varying motion patterns by clearing irrelevant entries and retaining only relevant historical motion vectors.
Data Source
AI summary
In a method of video decoding performed by a decoder, a current picture is obtained from a coded video bitstream. The current picture is divided into a plurality of units and divided into a plurality of tiles. Each tile includes at least one unit of the plurality of units. A first current unit in a first tile of the plurality of tiles is decoded. A first HMVP buffer is updated with a motion vector of the first current unit that has been decoded. A position of the first current unit in the first tile of the plurality of tiles is determined. The first HMVP buffer is reset when the first current unit is located in a first column of the first tile.