Adaptive Transform Basis Functions for Visual Data Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video compression techniques face challenges in accurately estimating the covariance function for transform-based coding, particularly for directional intra prediction residuals, which affects energy compaction and requires complex computations or suffers from inaccuracies.
Innovation Solution
A computationally efficient model is used to estimate the covariance function based on the gradient of boundary data values, allowing for the computation of Karhunen-Loève transform basis functions and adaptive transforms that improve energy compaction by relating residual statistics to prediction inaccuracy and boundary gradients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex computations are used to estimate the covariance function, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent uses a simplified covariance function estimation model that computes the function using only boundary gradient information rather than full residual analysis. This 'cheap' estimation approach provides sufficient accuracy for transform basis function computation without requiring complex, computationally intensive methods, thus resolving the contradiction between precision and complexity
Solution Approach 2:
The patent extracts only the essential boundary gradient information from the residual data to estimate the covariance function, rather than using the complete residual block. This extraction of critical features maintains estimation accuracy while significantly reducing computational complexity
2Loss of energy
If a larger number of transform coefficients are used, then energy compaction improves, but loss of substance increases
Solution Approach 1:
The patent changes the parameters used for transform basis function computation by using simplified covariance estimation based on boundary gradients. This parameter change enables effective energy compaction with fewer coefficients by optimizing the transform to better match the actual residual statistics without requiring a large number of coefficients
3Adaptability or versatility
If adaptive transforms are computed for each block, then adaptability improves, but productivity decreases
Solution Approach 1:
The patent segments the adaptability requirement by computing transform basis functions only where needed based on boundary gradient analysis, rather than performing full adaptive computation for every block. This selective segmentation maintains adaptability for blocks that benefit from it while improving overall encoding productivity
Data Source
AI summary
Encoding data includes: encoding a residual of a first portion of an array of data to generate a first set of coefficients; decoding the first set of coefficients to generate a decoded representation of the first portion; computing an estimated covariance function for a residual of a second portion of the array of data based on a model that includes a gradient of a plurality of boundary data values located on a boundary of the decoded representation of the first portion; computing a set of transform basis functions from the estimated covariance function; and encoding the residual of the second portion using a first transform that uses the computed set of transform basis functions.