Image Encoding Quantization Control via Threshold Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding methods, such as H.264, face challenges in efficiently controlling quantization values due to high computational complexity, making it difficult to adaptively assign rates based on nonlinear block characteristics.
Innovation Solution
An image encoding apparatus that calculates evaluation values for blocks, generates identification information by comparing these values with threshold values, and controls quantization parameters to enhance subjective image quality by adjusting quantization values based on visual characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple evaluation values are multiplied to calculate the degree of difficulty of encoding, then rate assignment control is achieved, but calculation complexity increases
Solution Approach 1:
The patent segments the continuous multiplication calculation into discrete threshold comparisons. Instead of computing a continuous function f(x1, x2, ..., xn) = x1 × x2 × ... × xn, the invention divides the evaluation space into multiple regions by comparing each evaluation value against predetermined thresholds, creating segmented decision regions that are computationally simpler.
Solution Approach 2:
The patent replaces complex continuous mathematical operations with simple threshold comparison operations that are computationally inexpensive. Each comparison operation is a 'cheap' discrete operation that can be executed efficiently, replacing the more expensive continuous multiplication and function evaluation operations.
2Ease of manufacture
If linear conversion of evaluation index is used, then mathematical formulation is simplified, but nonlinear block characteristics cannot be identified
Solution Approach 1:
The patent transitions from a one-dimensional linear evaluation approach to a multi-dimensional evaluation space by incorporating multiple evaluation values (luminance average, color difference average, luminance dispersion, motion vector rate). Each evaluation value becomes a dimension, and threshold comparisons create hyperplanar divisions in this multi-dimensional space, enabling nonlinear characteristic identification through the combined effect of multiple dimensions.
Solution Approach 2:
The patent applies different quantization strategies to different regions of the evaluation space. By dividing the multi-dimensional evaluation space into distinct regions using threshold comparisons, the system can assign different quantization parameters to blocks falling into different regions, adapting to local characteristics of each block rather than applying a uniform linear conversion.
3Manufacturing precision
If detailed rate assignment is implemented, then subjective image quality improves, but computational cost increases
Solution Approach 1:
The patent segments the evaluation space into multiple discrete regions using threshold comparisons, allowing detailed rate assignment to different block types while avoiding complex continuous calculations. Each region can be assigned a specific quantization strategy, achieving detailed control through discrete segmentation rather than continuous computation.
Data Source
AI summary
An image encoding apparatus includes an encoding unit which encodes an input image of a block, a calculating unit which calculates a plurality of evaluation values for the block, an identifying unit which generates identification information by comparing a plurality of evaluation values calculated by the calculating unit with a plurality of threshold values, and a controller which controls a quantization parameter on the basis of the identification information identified by the identifying unit.