Bi-Prediction Optical Flow Inter-Prediction Method
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution, high-quality images such as HD and UHD images leads to increased data transmission and storage costs due to the high amount of information required, necessitating a more efficient image compression technique.
Innovation Solution
The method and apparatus enhance inter-prediction efficiency by applying bi-prediction optical flow (BIO) prediction to derive sample-unit motion vectors, adaptively determining whether to apply BIO prediction to reduce computational complexity and improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If BIO prediction is applied to derive sample-unit motion vectors, then inter-prediction efficiency and coding performance are improved, but computational complexity increases
Solution Approach 1:
The patent applies BIO prediction dynamically and adaptively for specific blocks or regions where motion is complex, rather than uniformly across the entire image. The system determines whether to apply BIO prediction based on local motion characteristics, making the computational process flexible and adaptive to different content types, thereby improving prediction accuracy only where necessary while controlling overall complexity.
Solution Approach 2:
The patent implements sample-unit motion vector derivation through BIO prediction at the block or region level rather than globally. By applying the computationally intensive BIO algorithm only to specific areas with complex motion patterns and using simpler prediction methods for other areas, the system optimizes the balance between prediction accuracy and computational load.
2Measurement precision
If high-resolution image data is transmitted or stored, then image quality is improved, but transmission cost and storage cost increase
Solution Approach 1:
The patent extracts and transmits only the essential residual information after BIO prediction rather than the complete high-resolution image data. By deriving sample-unit motion vectors and generating prediction samples through BIO, the system identifies and transmits only the differences (residuals) between predicted and actual values, significantly reducing the data quantity required for high-quality reconstruction.
Solution Approach 2:
The patent transforms the representation of image data from raw pixel values to prediction residuals based on BIO-derived motion vectors. This parameter transformation allows high-resolution image quality to be maintained through accurate motion compensation while reducing the amount of data that needs to be transmitted or stored, as residuals require fewer bits than original high-resolution data.
Data Source
AI summary
An image decoding method performed by a decoding apparatus according to the present invention is characterised by comprising the steps of: deriving a reference picture list 0 (L0) and a reference picture list 1 (L1); deriving two motion vectors (MV) for a current block, wherein the two MVs include MVL0 for the L0 and MVL1 for the L1; determining whether to apply BIO prediction that derives motion vectors for each sample in the current block based on the MVL0 and the MVL1; deriving motion vectors for each sample in the current block based on the MVL0 and the MVL1 if it is determined to apply the BIO prediction; and deriving a prediction sample based on the motion vectors for each sample.