Constrained Soft-Decision Quantization for Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video encoding technologies, such as the ITU-T H.264/AVC standard, face computational challenges due to the high complexity of soft-decision quantization processes, which are computationally demanding because they require evaluating all states in the trellis for rate-distortion optimization.
Innovation Solution
Constrained soft-decision quantization is introduced, where the search range for quantization levels is limited to a defined neighborhood of hard-decision quantization levels, reducing the number of states and connections in the trellis, thereby decreasing computational complexity. This is achieved by setting a threshold value or dynamically adjusting the search range based on coefficient position, magnitude, or accumulated distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If soft-decision quantization is used to improve rate-distortion performance, then coding efficiency is improved, but computational complexity increases due to evaluating all states in the trellis
Solution Approach 1:
The patent applies partial action by evaluating only a subset of trellis states rather than all states. Specifically, it uses a constrained search that limits the number of states evaluated at each trellis node, thereby reducing computational complexity while maintaining most of the rate-distortion performance benefits of full soft-decision quantization.
Solution Approach 2:
The patent changes the parameter of search range in the trellis evaluation. By dynamically adjusting the search range based on rate-distortion cost thresholds, it transforms the computational complexity parameter while preserving the essential functionality of soft-decision quantization.
2Productivity
If the search range for soft-decision quantization is reduced to decrease computational complexity, then encoding speed is improved, but rate-distortion performance may deteriorate
Solution Approach 1:
The patent implements feedback by using the rate-distortion cost as a threshold to dynamically control the search range. The encoder evaluates states in order of increasing cost and stops when the cost exceeds the threshold, thereby adapting the search range to maintain performance while improving speed.
Solution Approach 2:
The patent makes the search range dynamic rather than fixed. The search range adapts based on the rate-distortion cost threshold and the specific characteristics of each coefficient, allowing the encoder to optimize between speed and performance on a per-coefficient basis.
Data Source
AI summary
A method of encoding a video using constrained soft-decision quantization. The soft-decision quantization includes first performing hard-decision quantization to obtain hard quantized coefficients and, then, obtaining a soft quantized coefficient using a rate-distortion calculation over a search range of quantization levels for a transform domain coefficient, wherein the search range of quantization levels for that transform domain coefficient is constrained within a number of quantization levels of a corresponding hard quantized coefficient. The search range may be based upon a fixed threshold, the coefficient position, the hard quantized coefficient magnitude, a threshold value less accumulated distortion, or other factors, including combinations of these factors. The accumulated distortion may be measured by an L1 norm.