Video Transform Index Prediction via Sign Reversal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies, such as HEVC, face limitations in flexibility and efficiency for predicting coding modes and transform selection, particularly in handling rectangular prediction units and transform units, which affects compression performance and computational complexity.
Innovation Solution
The proposed method involves binarizing a transform index to identify non-zero transform coefficients and reversing their signs based on specific bin values, allowing for more efficient transform selection and coding by representing the transform index with a reverse pattern of sign values, thereby reducing computational complexity and improving compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If transform index is explicitly signaled in bitstream, then transform selection accuracy is improved, but bitstream complexity and transmission overhead increase
Solution Approach 1:
Instead of directly signaling the transform index, the patent inverts the approach by signaling the difference between the predicted transform index and the actual transform index. This inversion allows the decoder to reconstruct the correct transform index while transmitting fewer bits, especially when the difference is small or zero.
Solution Approach 2:
The patent performs preliminary transform index prediction at both encoder and decoder sides using available information (neighboring blocks, prediction units, etc.) before actual encoding/decoding. This preliminary action reduces the information that needs to be transmitted in the bitstream, as only the correction to the prediction needs to be signaled.
2Adaptability or versatility
If multiple transform types are supported, then transform flexibility is improved, but computational complexity increases
Solution Approach 1:
The patent segments the transform selection process into two parts: a predicted transform index derived from neighboring blocks and prediction units, and a correction value signaled in the bitstream. This segmentation allows the majority of cases to be handled by simple prediction, reserving complex transform processing only for cases where the correction indicates a different transform type.
Solution Approach 2:
The patent applies full transform processing only when necessary (when the correction value indicates a different transform type), otherwise using the predicted transform index directly. This partial action approach reduces computational complexity by avoiding unnecessary transform calculations while maintaining flexibility when needed.
3Productivity
If transform coefficient processing is optimized, then compression efficiency is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary identification of non-zero transform coefficients and their sign patterns before actual transform processing. This preliminary action allows for optimized processing paths to be selected, where simple cases (e.g., all positive coefficients) can be handled more quickly while still achieving good compression efficiency.
Solution Approach 2:
The patent changes the parameter representation by using sign-value pairs for transform coefficients instead of traditional magnitude-phase representation. This parameter change enables more efficient processing and signaling, particularly when combined with the transform index prediction mechanism, improving the trade-off between compression efficiency and processing time.
Data Source
AI summary
A method of and an apparatus for controlling intra and/or inter prediction for decoding of a video sequence are provided. The method includes receiving a transform index which indicates a transform selection used for decoding an image block; binarizing a transform index using a preset number of bins; for each of a selected set of bins of the binarized transform index, identifying a group of non-zero transform coefficients in a transform coefficient block; and based on a value of a first bin of the selected set of bins of the binarized transform index and a value of a second bin of the selected set of bins of the binarized transform index, reversing a sign of the identified group of non-zero transform coefficients in the transform coefficient block.