ROI-Based Image Encoding Reducing Data Volume
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding/decoding technologies struggle to efficiently compress and decode images for both human vision and machine vision applications, such as surveillance and intelligent transportation, while maintaining high recognition accuracy.
Innovation Solution
The method involves preprocessing input images to identify and convert regions of interest (ROI), encoding the converted images, and providing metadata for inverse-conversion to the original image position, thereby reducing data volume and enhancing compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional video encoding/decoding technology is used, then image quality is maintained for human vision, but data volume is large and compression efficiency is insufficient for machine vision applications
Solution Approach 1:
The image is divided into multiple regions of interest (ROIs) based on their importance levels. Different encoding parameters and compression ratios are applied to different ROI segments, allowing critical regions to maintain high quality while less critical regions use higher compression, thereby reducing overall data volume while maintaining machine vision performance
Solution Approach 2:
Different quality levels and encoding strategies are applied to different spatial locations within the image based on their importance. High-importance ROIs receive higher quality encoding while low-importance regions use lower quality encoding, achieving local optimization of compression efficiency without sacrificing critical information
2Quantity of substance
If image compression is increased to reduce data volume, then transmission efficiency improves, but recognition accuracy for machine vision tasks deteriorates
Solution Approach 1:
Regions of interest are identified and marked before the encoding process begins. This preliminary segmentation allows the encoder to prioritize these regions and apply appropriate compression strategies that preserve recognition accuracy while reducing overall data volume
Solution Approach 2:
Encoding parameters such as quantization step size, transformation block size, and prediction mode are dynamically adjusted based on the importance level of different image regions. Critical ROIs use parameters that preserve detail for accurate recognition, while non-critical regions use aggressive parameter settings for maximum compression
3Reliability
If the entire image is encoded at high quality, then image quality is maintained, but data volume increases and compression efficiency decreases
Solution Approach 1:
The image is segmented into multiple importance-based regions, allowing differential quality treatment. High-importance ROIs are encoded at high quality to maintain reliability, while low-importance regions use lower quality encoding, significantly reducing overall data volume
Solution Approach 2:
Different quality levels are applied locally to different image regions based on their importance. This local quality optimization ensures that critical regions maintain high reliability while non-critical regions contribute minimally to data volume
Data Source
AI summary
A image encoding method according to present disclosure may comprise encoding an image based on a region of interest (ROI) including setting a ROI group, including the region of interest, in the image; converting the image based on the ROI group; and encoding a converted image. Here, the converted image may represent an image that a position of the ROI group is moved or a copped image generated to comprise the ROI group in the image.