Dual-Model Image Coding to Preserve Evidentiary Value

Solution Overview

Problem

AI-based image codecs face skepticism regarding the evidentiary value of restored images due to potential contamination with training data or randomness, limiting their use in law-enforcement applications like surveillance and monitoring.

Innovation Solution

An image encoding method that discriminates between different image regions, using a non-generative model for regions of interest (ROI) and a generative model for the remainder, ensuring decoding without reliance on extraneous information, thereby maintaining evidentiary value.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If AI-based generative image codecs are used to achieve high compression ratios, then image quality is improved, but evidentiary value deteriorates due to potential contamination with training data or randomized information

Engineering Contradiction:

Improvecompression ratioVSAvoidevidentiary value

Core Design Contradiction:

Loss of energyVSReliability

Solution Approach 1:

The image is divided into multiple regions, with at least one region encoded using a non-generative image model to preserve evidentiary value, while other regions may use generative models for higher compression. This segmentation allows simultaneous achievement of high compression ratios and maintained evidentiary value in critical regions.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different encoding models are applied to different regions of the image based on local requirements. Regions requiring evidentiary value (e.g., containing objects of forensic relevance) use non-generative models, while other regions use generative models for optimal compression, creating local quality variations in the encoding approach.

Inventive Principle:

Principle #3Local quality

2Reliability

If non-generative image models are used to maintain evidentiary value, then reliability is improved, but compression ratio deteriorates

Engineering Contradiction:

Improveevidentiary valueVSAvoidcompression ratio

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The image is divided into multiple regions, with at least one region encoded using a non-generative image model to preserve evidentiary value, while other regions may use generative models for higher compression. This segmentation allows simultaneous achievement of high compression ratios and maintained evidentiary value in critical regions.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different encoding models are applied to different regions of the image based on local requirements. Regions requiring evidentiary value (e.g., containing objects of forensic relevance) use non-generative models, while other regions use generative models for optimal compression, creating local quality variations in the encoding approach.

Inventive Principle:

Principle #3Local quality

3Reliability

If dual image models are used to balance compression and evidentiary value, then device complexity is increased, but this enables simultaneous achievement of compression ratios and evidentiary value preservation

Engineering Contradiction:

Improveevidentiary valueVSAvoidencoding system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The image is divided into multiple regions, with at least one region encoded using a non-generative image model to preserve evidentiary value, while other regions may use generative models for higher compression. This segmentation allows simultaneous achievement of high compression ratios and maintained evidentiary value in critical regions.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different encoding models are applied to different regions of the image based on local requirements. Regions requiring evidentiary value (e.g., containing objects of forensic relevance) use non-generative models, while other regions use generative models for optimal compression, creating local quality variations in the encoding approach.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS12401801B2System and method for image coding using dual image models

Publication Date: 2025.08.26 AXIS

AI summary

A method of encoding an image comprises establishing whether objects constituting one or more predefined object types or performing one or more predefined event types are visible in the image; in response to establishing that the objects are visible, encoding at least one region-of-interest of the image using a non-generative image model, thereby obtaining first image data; and encoding any remainder of the image using a generative image model, thereby obtaining second image data, wherein use of the non-generative image model enables decoding of the first image data without relying on information derived from images other than the encoded image or, if the image is a frame in a video sequence, enables decoding of the first image data without relying on information derived from images outside the video sequence.