Hybrid Image Encoding for ROI Accuracy and Low Bitrate
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI-based image compression methods, such as GANs, can introduce 'hallucinations' or incorrect features in encoded images, which may compromise their evidentiary value in applications requiring high accuracy, such as forensic investigations.
Innovation Solution
A method and system that combines generative and non-generative image models to encode images, where regions of interest with significant differences are re-encoded using non-generative models, ensuring high-quality encoding by comparing reference points in the original and decoded images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If AI-based generative models are used for image compression, then bitrate is significantly reduced and image quality is improved, but hallucinations or incorrect features are introduced in the encoded images
Solution Approach 1:
The image is divided into regions of interest and non-regions of interest. Different encoding strategies are applied to each segment: generative models for non-critical areas and conventional models for critical areas, thereby resolving the contradiction between compression efficiency and reliability
Solution Approach 2:
Different quality levels are applied to different parts of the image based on their importance. Regions of interest receive higher quality encoding with conventional models to ensure reliability, while non-critical areas use generative models for better compression, achieving local optimization of the contradiction
2Manufacturing precision
If generative models are used for image encoding, then image quality is visually improved, but incorrect features or details are produced
Solution Approach 1:
The encoding approach is made dynamic and adaptive based on the content being encoded. The system automatically switches between generative and conventional models depending on the region type and importance, allowing flexibility in balancing visual quality and feature accuracy
Solution Approach 2:
A quality assessment mechanism acts as an intermediary between the generative encoding process and the final output. This intermediary evaluates encoded regions and determines whether conventional re-encoding is needed, mediating between the visual quality benefits of generative models and the feature accuracy requirements
3Reliability
If conventional compression methods are used, then encoding reliability is maintained, but bitrate reduction and visual quality improvement are limited
Solution Approach 1:
The patent merges two different encoding approaches (generative and conventional models) into a unified hybrid system. This combination allows the system to leverage the bitrate efficiency of generative models while maintaining the reliability of conventional models in critical areas, resolving the contradiction between these two opposing requirements
Data Source
AI summary
A method of encoding images in a video, comprises: acquiring an original image from an image sensor of a video camera; encoding the original image using a generative image model, thereby obtaining a first encoded image; decoding the image to obtain a first decoded image; identify a region of interest (ROI) of the original image; for each ROI: perform an encoding quality check by comparing several reference points in the ROI of the original image against corresponding reference points in the ROI of the first decoded image, thereby obtaining a difference; if the difference is greater than a threshold, encoding the ROI using a non-generative image model, thereby obtaining a non-generative encoded image area; providing final encoded image data comprising a) the non-generative encoded image areas for the ROI having a difference greater than the threshold and b) the first encoded image for a remaining part of the original image.