Image Distortion Evaluation via Visual Texture Loss Entropy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating image distortion, such as analyzing pixel differences or model training, fail to accurately assess visual texture loss in enhanced images and are computationally complex.
Innovation Solution
An image distortion evaluation method that involves obtaining original and enhanced images, performing block processing, and calculating scale information entropy using a preset window size conforming to human visual characteristics to determine the degree of visual texture loss, reducing computational complexity without model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If pixel difference analysis is used to evaluate image distortion, then the evaluation process is simple, but it cannot reflect visual texture losses of enhanced images
Solution Approach 1:
The patent introduces information entropy as an intermediary metric to bridge the gap between simple pixel difference analysis and accurate visual texture loss detection. By using information entropy to quantify the texture characteristics of image blocks, the method achieves both computational simplicity and measurement precision, resolving the contradiction between ease of operation and measurement precision.
2Measurement precision
If model training is used to compute losses, then visual texture loss can be evaluated, but computational complexity is high
Solution Approach 1:
The patent replaces complex trained models with a lightweight information entropy calculation approach. Instead of using heavy computational resources for model training and inference, the method employs simple entropy calculations on image blocks, achieving comparable or better performance with significantly reduced computational complexity, especially suitable for resource-constrained devices like mobile phones.
Solution Approach 2:
The patent divides the image into multiple blocks and calculates information entropy for each block separately. This segmentation approach allows efficient parallel computation and reduces the overall computational burden compared to processing the entire image at once or using global models, thereby lowering device complexity while maintaining evaluation accuracy.
3Adaptability or versatility
If general distortion evaluation methods are used, then various types of distortion can be assessed, but visual texture loss specifically cannot be captured
Solution Approach 1:
The patent applies local quality analysis by dividing the image into blocks and calculating information entropy for each local region. This approach captures spatial variations in texture characteristics that general distortion metrics miss, enabling precise measurement of visual texture loss while maintaining adaptability to different image regions and enhancement types.
Data Source
AI summary
An image distortion evaluation method and apparatus, and a computer device. The method comprises: acquiring an original image and an enhanced image; respectively performing partitioning processing on the original image and the enhanced image, so as to obtain a plurality of first blocks of the original image and a plurality of second blocks of the enhanced image; acquiring a preset proportionate window size that accords with the characteristics of human eye vision, and according to the proportionate window size, respectively compiling statistics on first proportion information entropies respectively corresponding to the plurality of first blocks of the original image and second proportion information entropies respectively corresponding to the plurality of second blocks of the enhanced image; determining a visual texture loss degree of the enhanced image according to the first proportion information entropies corresponding to the first blocks and the second proportion information entropies corresponding to the second blocks.