Video Inter-Prediction Using Bi-Directional Optical Flow
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression techniques face challenges in achieving high encoding efficiency and image quality due to increasing picture size, resolution, and frame rate, necessitating improved inter-prediction methods.
Innovation Solution
A coding tool for inter-prediction that compensates for various motions of objects by using bi-directional optical flow and weighted bi-prediction, generating prediction blocks from multiple reference pictures with different weight assignments for luma and chroma components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional inter-prediction is used for video encoding, then encoding process is simple, but compression performance is insufficient for high resolution and high frame rate videos
Solution Approach 1:
The patent divides the prediction process into two separate stages: first generating a preliminary prediction block using traditional inter-prediction, then generating a refinement block using bi-directional optical flow. This segmentation allows the system to achieve high compression performance through the refinement stage while keeping the initial encoding process relatively simple.
Solution Approach 2:
The patent performs preliminary prediction using traditional inter-prediction methods before applying the more complex bi-directional optical flow refinement. This preliminary action provides a base prediction that reduces the amount of data needing refinement, thereby improving overall compression performance while managing encoding complexity through staged processing.
2Measurement precision
If bi-directional optical flow is applied to all blocks, then prediction accuracy is maximized, but encoding complexity and computational load increase significantly
Solution Approach 1:
The patent applies bi-directional optical flow refinement selectively to specific blocks rather than uniformly to all blocks. The refinement process is applied based on local characteristics of each block, such as motion complexity and prediction error, thereby maximizing prediction accuracy where needed while reducing computational complexity in areas where simple inter-prediction suffices.
Solution Approach 2:
The patent applies refinement action partially to only those blocks that benefit most from it, rather than applying full refinement to all blocks. This partial action approach ensures high prediction accuracy for difficult-to-predict blocks while avoiding the excessive computational load that would result from universal application of the refinement process.
3Measurement precision
If separate luma and chroma weight assignments are used, then prediction precision for different color components is improved, but processing complexity increases
Solution Approach 1:
The patent applies different weight assignments to luma and chroma components based on their respective characteristics. Luma weights are optimized for brightness information while chroma weights are optimized for color information, allowing each component to achieve maximum prediction precision with weights tailored to its specific properties rather than using a unified weighting approach.
Data Source
AI summary
A video decoding method, a video encoding method, and a method of transmitting a bitstream containing encoded video data are disclosed. The methods include determining first and second reference pictures and first and second motion vectors for bi-prediction by decoding a bitstream; generating a first reference block from the first reference picture referenced by the first motion vector and generating a second reference block from the second reference picture referenced by the second motion vector; and generating a prediction block of the target block using the first and second reference blocks. The methods also include generating a residual block of the target block and reconstructing the target block based on the prediction block of the target block and the residual block of the target block.