Mixed-Attribute Image Compression via Adaptive Block Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression methods, such as JPEG, struggle to maintain high resolution and gradation when images with different attributes, like texts, graphics, and photographic images, are mixed, leading to image quality deterioration.
Innovation Solution
An image processing apparatus that performs compression and extension processing based on the attribute of each pixel, using BTC compression for photographic images and degeneracy compression for texts and graphics, and vice versa during extension processing, to maintain gradation and resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If JPEG compression is applied to the entire image, then compression efficiency is improved, but image quality in text and graphic regions deteriorates due to loss of resolution
Solution Approach 1:
The image is divided into multiple blocks, and each block is independently classified and compressed using the most suitable method for its content type. This segmentation allows different compression strategies to be applied to different regions, maintaining high resolution in text/graphic blocks while achieving efficient compression in photographic blocks.
Solution Approach 2:
Different compression methods are applied to different regions based on their local characteristics. Text and graphic regions use degeneracy coding to preserve sharp edges and high resolution, while photographic regions use JPEG compression to achieve efficient compression with acceptable quality. This local quality approach ensures each region receives the appropriate treatment.
2Device complexity
If the entire image is compressed using a single compression method, then device complexity is reduced, but image quality deteriorates when mixed regions are present
Solution Approach 1:
The compression method is dynamically selected for each block based on its content characteristics. The system automatically determines whether each block contains text, graphics, or photographic content and applies the most appropriate compression method accordingly. This dynamic adaptation allows the system to handle mixed-region images effectively without requiring manual intervention.
Solution Approach 2:
The image compression system performs self-classification of each block's content type and automatically selects the optimal compression method without external intervention. The block classification unit analyzes each block's characteristics and makes the decision which compression algorithm to apply, enabling the system to serve itself in optimizing compression quality.
3Productivity
If block size is increased to enhance compression efficiency, then productivity is improved, but image quality in text and graphic regions deteriorates when texts are included in picture blocks
Solution Approach 1:
The image is divided into smaller blocks that can be independently classified and processed. This fine-grained segmentation ensures that text regions within picture blocks are separated into their own blocks, allowing degeneracy coding to be applied specifically to text regions while JPEG compression is applied to photographic regions, thereby preserving text resolution while maintaining compression efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed is an image processing apparatus comprising an image compression conversion unit to quantize an image having attribute data for each pixel, wherein when a region has the attribute data of a photographic image, the image compression conversion unit quantizes the region by a BTC method, wherein when a region has the attribute data of other than the photographic image, and the region is a halftone region, the image compression conversion unit quantizes the region by the BTC method, and wherein when a region has the attribute data of other than the photographic image, and the region is a high resolution region, the image compression conversion unit generates a density pattern for the region to quantize the region according to the generated density pattern.