Buffered Transform Signal Approximation for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in achieving a good rate-distortion relation while maintaining low buffer requirements, especially in FPGA implementations where bandwidth limitations restrict the size of the frame buffer needed for temporal prediction.
Innovation Solution
The proposed solution involves deriving a residual transform signal from a data stream by combining it with a buffered transform signal approximation of a previous picture, subjecting it to inverse spatial decorrelation, and encoding the residual signal, thereby reducing the need for full precision buffering and improving the rate-distortion relation with moderate buffer sizes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If a full sized frame buffer is used for temporal prediction, then the rate-distortion relation is improved, but the buffer requirements and bandwidth increase
Solution Approach 1:
The patent applies partial action by using only a subset of transform coefficients (e.g., only significant coefficients above a threshold) from the previous picture's transform signal, rather than buffering the complete transform signal. This selective buffering reduces the buffer size while maintaining adequate temporal prediction accuracy for the rate-distortion relation.
Solution Approach 2:
The patent extracts only the necessary components (significant transform coefficients) from the previous picture's transform signal for temporal prediction, discarding less important coefficients. This extraction approach reduces buffer requirements while preserving the essential information needed for maintaining good rate-distortion performance.
2Measurement precision
If a full sized frame buffer is used for temporal prediction, then the temporal prediction accuracy is improved, but the bandwidth requirements increase
Solution Approach 1:
The patent extracts only the significant transform coefficients that contribute meaningfully to temporal prediction accuracy, eliminating the need to transport less important coefficients. This reduces bandwidth requirements while maintaining adequate temporal prediction accuracy.
Solution Approach 2:
The patent uses partial action by transmitting only a subset of transform coefficients (those above a significance threshold) for temporal prediction, rather than all coefficients. This partial transmission reduces bandwidth while preserving sufficient prediction accuracy.
3Quantity of substance
If approximations of transform coefficients are used in the buffer, then the buffer size is reduced, but the manufacturing precision of the transform signal decreases
Solution Approach 1:
The patent applies local quality by using full precision for significant transform coefficients (those above a threshold) while using approximations or zero for less significant coefficients. This selective precision approach reduces buffer size while maintaining transform signal precision where it matters most for image quality.
Solution Approach 2:
The patent changes the precision parameter of transform coefficients based on their significance. Significant coefficients are maintained at full precision, while less significant coefficients are approximated or set to zero, optimizing the trade-off between buffer size and transform signal precision.
Data Source
AI summary
An apparatus for decoding a sequence of pictures from a data stream is configured for decoding a picture of the sequence by deriving a residual transform signal of the picture from the data stream; combining a residual transform signal with a buffered transform signal approximation of a previous picture of the sequence so as to obtain a transform signal representing the picture, the transform signal having a plurality of transform coefficients; and subjecting the transform signal to a spectral-to-spatial transformation. The apparatus is configured for deriving the buffered transform signal approximation from a further transform signal representing the previous picture so that the buffered transform signal approximation has approximations of further transform coefficients of the further transform signal.