An Objective Image Quality Evaluation Method Based on Manifold Feature Similarity

Abstract

The invention discloses an objective evaluation method of image quality based on the similarity of manifold features, which first adopts two strategies of visual salience and visual threshold to remove image blocks that are not important to visual perception, that is, image block rough selection and fine selection process After completing the block selection, use the best mapping matrix to extract the manifold feature vector of the image block selected from the original undistorted natural scene image and the distorted image to be evaluated, and then measure the distorted image by the similarity of the manifold feature Structural distortion; after considering the impact of image brightness changes on the human eye, the brightness distortion of the distorted image is calculated based on the average value of the image block; finally, the quality score is obtained according to the structural distortion and brightness distortion, which makes this method have higher evaluation accuracy , also expands its ability to evaluate various types of distortion, and the evaluation performance is not affected by the image content and distortion type, and has a high consistency with the subjective perception quality of the human eye.

Description

technical field [0001] The invention relates to an image quality evaluation method, in particular to an objective image quality evaluation method based on manifold feature similarity. Background technique [0002] Quantitative evaluation of image quality is a challenging problem in the field of image processing. Since humans are the ultimate recipients when viewing images, image quality assessment methods should be as effective in predicting perceived visual quality as humans. Although the traditional image quality evaluation methods based on fidelity such as Peak Signal-to-Noise Ratio (PSNR) can better evaluate the image quality with the same content and distortion, but in the face of multiple images and multiple When there is such a distortion, the evaluation result is far from the subjective perception. The purpose of the perceptual quality evaluation method is to obtain an evaluation result with high consistency with the visual perception quality by simulating the over...

AI Technical Summary

Problems solved by technology

The above structure-based image quality evaluation methods all obtain image quality from structural information such as the edge and contrast of the image, while the image quality evaluation methods designed from the characteristics of the human visual system mainly focus on the ability of the human eye to detect distortion. The evaluation of image quality is based on the nonlinear geometric structure of the image and the perception of the human eye; but some studies have shown that for visual perception phenomena, the manifold is the basis of perception, and the brain uses the manifold Perceive things, and natural scene images usually contain manifold structures, which have the nature of manifold nonlinearity

Therefore, traditional image quality evaluation methods cannot obtain objective evaluation results that are consistent with subjective perception quality

Images