Image Coding Using Scene Quality Distribution and Adaptive Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image coding methods struggle to efficiently differentiate between region of interest (ROI) and background regions, leading to visible boundaries and reduced coding efficiency due to uniform bit allocation and limited quantization parameter adjustments.
Innovation Solution
An image coding and decoding method that models image quality distribution across scenes, determining unique quantization parameters for each region to gradually lower image quality from ROI to background, using adaptive quantization and entropy coding, and divides images into rectangular slices for independent coding and decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If uniform quantization parameter is used across the whole image, then device complexity is reduced, but image quality in region of interest deteriorates
Solution Approach 1:
The patent applies local quality by dividing the image into multiple regions (first region and second region) and assigning different quantization parameters to each region. The first region (containing ROI) uses a first quantization parameter while the second region (background) uses a second quantization parameter, allowing differentiated quality control matching human visual characteristics without excessive complexity
Solution Approach 2:
The patent segments the image into distinct regions based on importance (ROI-containing region vs. background region) and applies separate quantization strategies to each segment. This segmentation enables targeted quality optimization while maintaining manageable system complexity through region-based processing
2Manufacturing precision
If different quantization parameters are assigned to each unit block, then image quality in region of interest is improved, but device complexity increases
Solution Approach 1:
Instead of assigning unique quantization parameters to every unit block, the patent applies local quality by defining two distinct regions and assigning one quantization parameter to each region. This approach maintains image quality in the ROI while avoiding the excessive complexity of per-block parameter management
Solution Approach 2:
The patent applies partial action by not using different quantization parameters for every single block, but rather for two strategically defined regions. This partial differentiation achieves the necessary quality improvement in ROI areas without the excessive complexity of complete per-block customization
3Manufacturing precision
If object separation is performed for each object, then image quality in region of interest is improved, but device complexity increases
Solution Approach 1:
The patent segments the image into two functional regions (first region containing ROI, second region as background) rather than separating individual objects. This region-based segmentation achieves quality differentiation while avoiding the complex object separation and shape coding processes
Solution Approach 2:
The patent applies local quality through region-based quantization parameters rather than object-based approaches. By defining a first region for ROI and a second region for background with different quantization parameters, it achieves localized quality control without the complexity of object identification and separation
4Device complexity
If rectangular slices are used to divide the image, then device complexity is reduced, but adaptability to arbitrary region shapes deteriorates
Solution Approach 1:
The patent segments the image into rectangular slices with a predetermined number of lines, providing a simple and manageable division structure. This segmentation approach reduces device complexity while still enabling effective region-based processing through the rectangular slice organization
Data Source
AI summary
An image coding method and apparatus considering human visual characteristics are provided. The image coding method comprises (a) modeling image quality distribution of an input image in units of scenes such that the quality of an image input in units of scenes is gradually lowered from a region of interest to a background region, (b) determining a quantization parameter of each region constituting one scene according to the result of modeling of image quality distribution, (c) quantizing image data in accordance with the quantization parameter, and (d) coding entropy of the quantized image data.