Semantic Relevance-Based Image Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression techniques apply a uniform level of compression to entire images, potentially overcompressing important content and undercompressing unimportant content, without considering semantic relevance.
Innovation Solution
A method that identifies semantically relevant portions of an image and applies differential compression, using techniques such as face-tracking, object recognition, and autofocus data to quantify image information differently based on relevance, allowing for more data retention in areas of interest and greater data discard in less relevant areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If uniform compression is applied to the entire image, then compression efficiency is improved, but image quality of important content deteriorates
Solution Approach 1:
The patent applies different compression qualities to different regions of the image based on semantic relevance. Semantically important regions (such as faces, objects of interest) are compressed with higher quality to preserve image quality, while less important regions are compressed with lower quality to improve overall compression efficiency. This resolves the contradiction by making compression quality spatially variable rather than uniform.
Solution Approach 2:
The patent segments the image into multiple regions based on semantic relevance analysis. By dividing the image into semantically important and less important regions, the system can apply differential compression strategies to each segment, thereby achieving both high compression efficiency and preserved image quality in critical areas.
2Productivity
If lossy compression is applied to discard maximum data, then compression efficiency is improved, but data retention of important content deteriorates
Solution Approach 1:
The patent applies varying levels of data discarding based on regional semantic importance. In semantically important regions, less data is discarded to maintain data retention, while in less important regions, more data is discarded to achieve higher compression efficiency. This creates a localized data retention strategy that balances overall compression performance.
Solution Approach 2:
By segmenting the image based on semantic relevance, the patent enables different data retention strategies for different segments. Important segments retain more original data with minimal lossy compression, while non-critical segments undergo more aggressive data discarding, thus resolving the contradiction between compression efficiency and data retention.
3Device complexity
If single level of compression is applied throughout the image, then device complexity is reduced, but compression efficiency deteriorates
Solution Approach 1:
The patent performs preliminary semantic relevance analysis and region segmentation before applying compression. By pre-identifying important and less important regions, the system prepares a compression map that guides subsequent differential compression. This preliminary action enables efficient multi-level compression without significantly increasing overall system complexity.
Solution Approach 2:
The patent introduces dynamic compression levels that adapt to the semantic content of different image regions. Rather than using a static single compression level, the system dynamically adjusts compression strength based on regional importance, thereby improving compression efficiency while managing complexity through adaptive algorithms.
Data Source
AI summary
A method includes: receiving an image in a system; identifying, using the system, a portion of the received image that is semantically relevant; and compressing, using the system, the received image based on the identified portion.