Video Coding Prediction Loop Filter Using Shearlet Transform
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video codecs lack efficient predictive coding and decoding methods, particularly in the transform domain, which hampers de-blocking and de-noising processes, and fails to adapt filtering to the shearlet domain.
Innovation Solution
The implementation of a shearlet domain-based predictive encoding and decoding system that applies a shearlet transform to images, classifies coefficients, and uses dual shearlet filters for re-transformation and summation to generate an in-loop filtered version, ensuring different classes of coefficients contribute differently to the filtered result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional block-based predictive coding is used, then device complexity is reduced, but manufacturing precision (coding efficiency) deteriorates due to insufficient de-blocking and de-noising capabilities
Solution Approach 1:
The patent segments the filtering process by classifying shearlet coefficients into different groups (e.g., edge coefficients vs. non-edge coefficients) and applying different filtering operations to each class. This allows selective application of complex filtering only where needed, improving coding efficiency while controlling overall complexity.
Solution Approach 2:
The patent transforms the filtering operation from the spatial domain to the shearlet transform domain. This dimensional change enables more effective de-blocking and de-noising by exploiting the directional selectivity and multi-scale properties of shearlet coefficients, thereby improving coding efficiency.
2Manufacturing precision
If adaptive loop filtering with multiple filter types is applied, then manufacturing precision (de-blocking performance) is improved, but device complexity increases due to multiple filtering operations
Solution Approach 1:
The patent applies different filtering strengths and types to different regions of the image based on local characteristics detected through shearlet coefficient analysis. Edge regions receive different treatment than smooth regions, optimizing de-blocking performance locally while avoiding unnecessary complexity in uniform areas.
Solution Approach 2:
The patent changes the parameters of the filtering operation by applying different threshold values, filter strengths, or filter types based on the classification of shearlet coefficients. This allows adaptive de-blocking performance while managing complexity through parameter adjustment rather than multiple independent filter systems.
3Manufacturing precision
If classification of shearlet coefficients is performed, then manufacturing precision (de-noising capability) is improved, but device complexity increases due to classification and selective processing
Solution Approach 1:
The patent replaces complex mechanical or algorithmic classification systems with a shearlet transform-based classification approach. The inherent directional and multi-scale properties of shearlet coefficients provide natural classification categories that simplify the filtering decision process while improving de-noising capability.
4Manufacturing precision
If in-loop filtering is applied to improve prediction accuracy, then manufacturing precision (prediction accuracy) is improved, but loss of time increases due to additional processing steps
Solution Approach 1:
The patent performs classification and filtering decisions in advance during the encoding process, storing the classification results and filter parameters. During decoding, these pre-computed parameters are directly applied without repeating the classification process, thereby improving prediction accuracy while minimizing additional processing time at decode.
Solution Approach 2:
The patent copies the filtered reference pictures and filter parameters from the encoder to the decoder through the bitstream. This allows both encoder and decoder to use the same filtering operations, improving prediction accuracy consistency while avoiding the need for the decoder to perform complex real-time filtering calculations.
Data Source
Figure 1
Figure 2
Figure 3a~3b
AI summary
A video decoder or encoder comprising a prediction loop filter, the prediction loop filter applying a shearlet transform with a set of shearlet filters onto at least a picture portion so as to obtain a set of coefficient matrices, subjecting coefficients of the set of coefficient matrices to classification so as to obtain a set of partial coefficient matrices per coefficient matrix, and deriving an in-loop filtered version of the picture portion by subjecting the set of partial coefficient matrices of the set of coefficient matrices to a shearlet retransform with a set of dual shearlet filters and summation, wherein, for each coefficient matrix, the partial coefficient matrices differently contribute to the in-loop filtered version.