Dynamic Transform Unit Grouping for Image Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding methods face inefficiencies in compressing images due to fixed transform unit sizes that do not adapt to changing image characteristics, leading to suboptimal compression ratios.
Innovation Solution
The method allows for a transform unit size larger than the prediction unit, enabling flexible grouping of prediction units into transform units for efficient frequency domain transformation, quantization, and entropy encoding, which adapts to varying image characteristics by determining optimal coding and transform unit sizes based on rate-distortion costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a fixed transform unit size is used, then the encoding process is simple, but the compression ratio is suboptimal due to inability to adapt to changing image characteristics
Solution Approach 1:
The transform unit size is made dynamic and adaptable rather than fixed. The encoder determines optimal transform unit sizes based on rate-distortion costs and image characteristics, allowing the system to adapt transform unit dimensions to match local image features such as texture complexity and gradient variations, thereby improving compression efficiency without excessive complexity
Solution Approach 2:
The patent changes the parameter of transform unit size from a fixed value to a variable that can be optimized based on rate-distortion analysis. By adjusting transform unit dimensions according to image characteristics and encoding goals, the system achieves better compression ratios while maintaining manageable encoding complexity through systematic parameter optimization
2Productivity
If transform unit size is larger than prediction unit size, then compression efficiency is improved, but the encoding complexity increases due to flexible grouping requirements
Solution Approach 1:
The patent segments the image into prediction units first, then groups multiple prediction units into larger transform units. This segmentation approach allows flexible grouping where multiple smaller prediction units can be combined into a single larger transform unit for frequency transformation, improving compression efficiency by better matching image characteristics while maintaining organized processing structure
Solution Approach 2:
Multiple prediction units are merged into a single transform unit when it is beneficial for compression. By combining adjacent prediction units into a larger transform unit, the system performs frequency transformation on a larger block that better captures image characteristics, thereby improving encoding efficiency while the merging process is guided by rate-distortion optimization to control complexity
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A method of encoding an image by transforming a group of prediction units into a frequency domain. A prediction unit may be predicted by using prediction values of the prediction units in the group of prediction units, instead of pixels generated by encoding and then restoring the other prediction units in the group of prediction units. Thus, an encoding compression ratio may be increased as a result of grouping the prediction units.