Affine Model Prediction for Video Decoding Memory Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding techniques face challenges in reducing memory requirements for affine motion vector prediction, which is essential for efficient video compression and decoding, especially in high-resolution video streams.
Innovation Solution
The proposed solution involves using affine model prediction based on motion information from neighboring blocks, where the processing circuitry identifies consecutive minimum compensation blocks to determine the parameters of the affine model and reconstruct samples, reducing the need to store affine model parameters of neighboring blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If affine model prediction is implemented using conventional methods, then video compression efficiency is improved, but memory requirements increase due to storing affine model parameters of neighboring blocks
Solution Approach 1:
The patent extracts only the essential motion information from neighboring blocks that is needed for affine prediction, rather than storing complete affine model parameters. By taking out only the necessary components (motion vectors and gradients) and discarding redundant data, the memory footprint is significantly reduced while maintaining compression efficiency.
Solution Approach 2:
The patent applies local quality by storing different types of information with different memory requirements based on their importance and usage frequency. Critical motion vectors are stored with full precision, while less critical affine parameters are stored with reduced precision or derived on-demand, optimizing the balance between memory usage and prediction accuracy.
2Measurement precision
If complete affine model parameters are stored for all neighboring blocks, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the affine model parameters into different categories based on their importance and usage. Instead of treating all parameters uniformly, it divides them into essential motion vectors that require full precision and secondary parameters that can be derived or stored with less detail, reducing overall system complexity.
Solution Approach 2:
The patent introduces intermediate calculations and derived parameters as mediators between stored data and final prediction results. Rather than storing and processing all raw parameters directly, it uses intermediate gradient calculations and motion vector differences to bridge the gap, simplifying the data structure while maintaining accuracy.
Data Source
AI summary
Aspects of the disclosure provide methods and apparatuses for video encoding/decoding. In some examples, an apparatus for video decoding includes receiving circuitry and processing circuitry. For example, the processing circuitry decodes prediction information of a current block in a current picture from a coded video bitstream. The prediction information is indicative of an affine based motion vector prediction. Further, the processing circuitry identifies consecutive minimum compensation blocks that are adjacent of the current block and affine coded. Then, the processing circuitry determines, based on motion information of at least two minimum compensation blocks in the consecutive minimum compensation blocks, parameters of an affine model, and reconstruct at least a sample of the current block based on the affine model that is used to transform between the current block and a reference block in a reference picture that has been reconstructed.