Grouped Transform Sign Prediction for Lower-Cycle Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for sign prediction in video coding/decoding, such as those described in NPL 1 and NPL 2, result in an exponential increase in calculation combinations and cycles due to the addition of sign prediction to quantization and transform processing.
Innovation Solution
A video decoding apparatus that predicts the sign of transform coefficients by deriving a cost with reference to neighboring block boundaries, divides coefficients into groups, and modifies signs using coefficient sign prediction errors, and includes parameters for non-separable transforms and scaling matrices to optimize sign prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sign prediction is added to quantization and transform processing, then prediction accuracy is improved, but the number of calculation cycles increases
Solution Approach 1:
The patent segments the transform coefficients into multiple groups based on their positions in the transform block. Different sign prediction methods are applied to different groups: the first group (e.g., coefficients with larger magnitudes or specific positions) uses full sign prediction to improve accuracy, while the second group uses simplified or skipped sign prediction to reduce calculation cycles. This segmentation allows the system to achieve good prediction accuracy for the most important coefficients while minimizing the overall computational burden.
2Adaptability or versatility
If the number of signs to be predicted increases, then prediction coverage is improved, but the amount of calculation increases exponentially
Solution Approach 1:
The patent applies different levels of sign prediction processing to different local regions (groups) of transform coefficients. The first group of coefficients, which typically has greater impact on reconstruction quality, receives comprehensive sign prediction processing. The second group receives reduced or no sign prediction processing. This local differentiation allows the system to concentrate computational resources where they provide the most benefit, avoiding exponential calculation increase while maintaining adequate prediction coverage.
Data Source
AI summary
Video coding and decoding apparatuses that can enhance coding efficiency are provided. A video decoding apparatus according to an aspect of the present invention includes a parameter decoder configured to decode a transform coefficient and a coefficient sign prediction error from coded data, and a sign prediction unit configured to predict a sign of the transform coefficient. The sign prediction unit derives a cost with reference to a boundary pixel with a neighboring block for a candidate of a combination of signs of transform coefficients without sign information, selects the sign, based on the cost, and modifies the predicted transform coefficient sign, using the coefficient sign prediction error decoded by the parameter decoder. The sign prediction unit divides the transform coefficients without the sign information into multiple groups, and performs the prediction on a transform coefficient of the transform coefficients for each of the multiple groups.