Image Compression Using Region-Specific Ratios for Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image compression methods struggle to accurately recognize objects that are inconspicuous to the human visual sense, leading to difficulties in object recognition, especially for small objects at a distance, such as cars, which is crucial for driving assistance systems.
Innovation Solution
An image compression apparatus and method that extracts a target region including objects of a predetermined size, compressing it at a lower ratio than non-target regions, allowing for accurate object recognition upon decompression, and optionally predicting target region movement to enhance processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional image compression methods are used, then high compression ratios can be achieved, but object recognition accuracy deteriorates for small and distant objects
Solution Approach 1:
The patent applies different compression ratios to different regions of the image based on their importance. Specifically, regions containing small objects (detected through object detection algorithms) are compressed at lower ratios to preserve recognition accuracy, while other regions are compressed at higher ratios. This local differentiation resolves the contradiction by maintaining high compression overall while protecting critical information.
Solution Approach 2:
The image is segmented into multiple regions based on object detection results. The patent divides the image into regions containing small objects and regions without such objects, then applies different compression strategies to each segment. This segmentation allows the system to achieve high compression ratios for non-critical regions while maintaining accuracy for regions containing important objects.
2Productivity
If uniform compression is applied to the entire image, then processing efficiency is improved, but object recognition accuracy for small objects deteriorates
Solution Approach 1:
The patent performs object detection and identifies regions containing small objects before applying compression. This preliminary action allows the compression algorithm to pre-determine which regions require special handling, enabling efficient processing while maintaining accuracy for critical regions. The preliminary detection step guides the subsequent differential compression process.
Solution Approach 2:
Different compression parameters and algorithms are applied to different regions of the image based on their content. Regions containing small objects use compression settings optimized for preserving fine details and small features, while other regions use more aggressive compression. This local quality approach maintains processing efficiency through automated region-based processing while ensuring accuracy where needed.
Data Source
AI summary
An image processing apparatus includes a target region extraction unit configured to extract from an image, a target region that is a region including an object having a predetermined size, and an image compression unit configured to compress the image on the basis of a result of extraction of the target region.