Video Coding Block Subdivision for Natural Motion Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression standards, such as MPEG and H.264, face limitations in accurately predicting natural motions like rotation and zoom due to their reliance on translation-based motion compensation, leading to suboptimal compression performance and increased residue data transmission.
Innovation Solution
A method that determines control vectors for blocks and divides them into sub-blocks to account for natural movements, using forward or backward motion compensation, and inserts reconstruction information into the signal to facilitate accurate prediction while maintaining low complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If translation-based motion compensation is used, then device complexity is reduced, but prediction quality for natural movements deteriorates
Solution Approach 1:
The patent divides a block into multiple sub-blocks and determines motion vectors for each sub-block independently. This segmentation allows the system to capture local variations in natural movements (rotation, zoom, shear) without requiring complex global transformation models, thus maintaining prediction quality while controlling device complexity.
Solution Approach 2:
The patent introduces dynamic motion compensation by allowing different motion vectors for different sub-blocks within the same block. This dynamic approach enables the system to adapt to varying motion patterns across different regions of a block, accurately representing natural movements without imposing excessive computational complexity.
2Measurement precision
If advanced motion compensation techniques are used to account for natural movements, then prediction quality improves, but the volume of residue information increases
Solution Approach 1:
By segmenting blocks into sub-blocks with individual motion vectors, the patent achieves better prediction quality for natural movements. The segmentation allows accurate modeling of complex motions while keeping each sub-block's residue information manageable, preventing overall residue volume from becoming excessive.
Solution Approach 2:
The patent applies local quality by determining motion vectors specific to each sub-block rather than using a single global motion vector. This local adaptation enables accurate prediction of natural movements in each region while minimizing the residue information that needs to be transmitted, as each sub-block's prediction is optimized for its local characteristics.
3Measurement precision
If multiple motion vectors per block are used, then prediction quality for natural movements improves, but device complexity increases
Solution Approach 1:
The patent manages device complexity by segmenting blocks into a limited number of sub-blocks (e.g., 4 sub-blocks per block), each with its own motion vector. This segmentation provides sufficient accuracy for natural movements while keeping the number of motion vectors manageable, avoiding excessive device complexity.
Solution Approach 2:
The patent uses partial action by applying motion compensation to only the most significant sub-blocks or using simplified motion vector derivation for some sub-blocks. This approach achieves adequate prediction quality for natural movements without processing excessive motion vectors, thereby controlling device complexity.
Data Source
Figure 1A~1B
Figure 2
Figure 3A~3F
AI summary
The invention relates to a method of coding a sequence of images, comprising a step of coding at least one block of a current image of the sequence, implementing a prediction (11) associating the block with at least one corresponding block in a reference image. According to the invention, the prediction (11) comprises the following steps, for at least one given block of the current image or of the reference image: determination (111) of a plurality of control vectors for the block; at least one splitting (112) of the block into sub-blocks; determination (113) of a plurality of control vectors for the sub-blocks; determination (114) of a prediction value for at least one point of the current image. Moreover, the method comprises a step of inserting, into a signal representative of the sequence, information regarding reconstruction of the control vectors for the given block.