Bi-Prediction Motion Vector Derivation for Interpolation Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding technologies fail to efficiently utilize motion vectors from multiple reference pictures for bi-prediction, leading to increased bit rates and inefficiencies in image encoding and decoding processes.
Innovation Solution
The proposed method and apparatus generate a current interpolation block using motion vectors from both a first and a second reference picture, reducing the need for explicit signaling of multiple motion vectors by deriving and selecting motion vector candidates from temporal neighboring blocks in these reference pictures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion vectors from multiple reference pictures are explicitly signaled for bi-prediction, then prediction accuracy is improved, but bit rate increases
Solution Approach 1:
The patent creates a virtual reference picture by copying and combining blocks from two different reference pictures using motion vectors. This virtual reference picture serves as an interpolated representation that captures motion information from multiple sources without explicitly signaling all motion vectors, thereby maintaining prediction accuracy while reducing bit rate
Solution Approach 2:
The patent introduces a virtual reference picture as an intermediary between the two actual reference pictures. This intermediary contains interpolated block information that mediates the prediction process, allowing the decoder to reconstruct the current block using combined motion information without receiving explicit signaling for multiple motion vectors
2Manufacturing precision
If multiple motion vectors are signaled for bi-prediction, then encoding precision is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary interpolation of blocks from reference pictures to create the virtual reference picture before the actual prediction process. This preliminary action pre-computes the combined motion information, simplifying subsequent decoding operations and reducing processing complexity while maintaining encoding precision
Solution Approach 2:
By copying and combining blocks from reference pictures into a virtual reference picture, the patent creates a pre-processed data structure that encapsulates multiple motion vectors. This copying approach simplifies the decoding process by replacing complex multi-vector processing with simpler reference to the pre-computed virtual picture
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image decoding method, including: obtaining a motion vector of at least one temporal neighboring block included in at least one of a first reference picture and a second reference picture; deriving one or more motion vector candidates from the motion vector of the at least one temporal neighboring block, wherein each motion vector candidate of the one or more motion vector candidates includes a first motion vector indicating a block within the first reference picture and a second motion vector indicating a block within the second reference picture; selecting a motion vector candidate for a current interpolation block from among the one or more motion vector candidates; generating the current interpolation block using a first block within the first reference picture indicated by the first motion vector included in the selected motion vector candidate and a second block within the second reference picture indicated by the second motion vector included in the selected motion vector candidate; and reconstructing a current block using the current interpolation block.