Spatially Enhanced Transform Coding for Video Prediction Errors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern video and image coding systems face suboptimal performance due to inefficient transform coding of prediction error signals, especially when the prediction error becomes less correlated, leading to increased distortion and bitrate in compressed video representations.
Innovation Solution
The prediction error signal is represented as a weighted sum of selected transform basis functions and quantized spatial samples, combining the strengths of both transform and spatial coding approaches to efficiently represent well-correlated and poorly correlated components of the signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If transform coding is used for prediction error signals, then compression efficiency is improved for well-correlated components, but performance deteriorates when prediction error becomes less correlated
Solution Approach 1:
The patent dynamically adapts the coding approach by selecting between transform coding and spatial coding based on the correlation characteristics of the prediction error signal. The system determines whether to apply transform coding or spatial coding on a block-by-block basis, allowing the coding method to adapt to local signal characteristics and maintain high performance across diverse content types.
Solution Approach 2:
The patent changes the coding parameter (transform coding vs. spatial coding) based on the statistical properties of the prediction error signal. By analyzing the correlation characteristics and selecting the appropriate coding mode accordingly, the system optimizes compression efficiency for both well-correlated and poorly correlated components.
2Measurement precision
If more transform coefficients are used to represent prediction error, then accuracy is improved, but bitrate increases
Solution Approach 1:
The patent extracts and separates well-correlated components (coded with transform coefficients) from poorly correlated components (coded with spatial samples). By taking out the poorly correlated components that do not benefit from transform coding and representing them directly in the spatial domain, the system achieves accurate representation without unnecessarily increasing bitrate.
Solution Approach 2:
The patent segments the prediction error signal into different components based on their correlation characteristics. Well-correlated components are processed with transform coding while poorly correlated components are processed with spatial coding, allowing each segment to be encoded with the most efficient method and avoiding uniform over-coding.
3Productivity
If transform coding is applied to all prediction error components, then well-correlated components are efficiently coded, but poorly correlated components like sensor noise and high-frequency textures are not effectively represented
Solution Approach 1:
The patent applies different coding qualities and methods to different parts of the prediction error signal based on local characteristics. Well-correlated components receive transform coding for efficiency, while poorly correlated components like sensor noise and high-frequency textures receive spatial coding for accurate representation, with each component treated according to its specific needs.
Data Source
AI summary
A system and method for improving prediction error coding performance of various video compression algorithms. Various embodiments combine the energy compaction features of transform coding with localization properties of spatial coding. In various embodiments, the effects of pixel “outliers” are removed from the transform and are coded separately as a spatially coded pixel prediction, thereby improving the coding gain of the transform.