Adaptive image quality objective evaluation method

Abstract

The invention discloses an adaptive image quality objective evaluation method. When the adaptive image quality objective evaluation method is adopted to obtain the structural similarity between two image blocks which have the same coordinate position in an original undistorted image and a distorted image to be evaluated, not only are the brightness mean value and standard deviation of all pixels in each image block in the original undistorted image and the distorted image to be evaluated, as well as the covariance among all the pixels in two image blocks in the original undistorted image and the distorted image to be evaluated utilized, but also the distortion types of the distorted image to be evaluated are utilized in a combined manner. Thus, the adaptive image quality objective evaluation method can determine the structural similarity between two image blocks according to the distortion types of the distorted image to be evaluated in the perspective of adaptive evaluation, and therefore, the consistency between an objective evaluation result and subjective perception about image quality can be improved.

Description

technical field [0001] The invention relates to an image quality evaluation technology, in particular to an adaptive image quality objective evaluation method. Background technique [0002] Image is an important way for human beings to obtain information, and the quality of image will directly affect the accuracy and integrity of information. However, in the process of image acquisition, processing, storage and transmission, it is inevitable to suffer different degrees of distortion. How to measure the loss of image information caused by these distortions has become a hot topic. The earliest and most widely used methods for objective evaluation of image quality are the peak signal-to-noise ratio method and the absolute difference sum method, but because they are all pixel-level methods, they do not take into account the relationship between pixels, so As a result, the objective evaluation results are often far from the visual perception of the human eye. As a result, two t...

AI Technical Summary

Problems solved by technology

However, in the process of image acquisition, processing, storage and transmission, it will inevitably be subject to different degrees of distortion. How to measure the loss of image information caused by these distortions has become a hot topic.

The earliest and most widely used methods for objective evaluation of image quality are the peak signal-to-noise ratio method and the absolute difference sum method, but because they are all pixel-level methods, they do not take into account the relationship between pixels, so As a result, the objective evaluation results are often far from the visual perception of the human eye

As a result, two categories of evaluation methods combined with the perceptual characteristics of the human eye have been proposed: the first category is based on the perception mechanism of the human visual system (HVS, human visual system); the second category is based on the structure similarity (SSIM, structure similarity ) to measure image quality, the former cannot accurately use mathematical models to simulate visual perception in the actual algorithm modeling due to the lack of understanding of the HVS perception mechanism, thus limiting the development of this type of method; the latter believes that the human eye is observing When using an image, the structural information of the image can be actively obtained, and the human eye also perceives the image from these structural information, so the quality of the distorted image can be evaluated by measuring the degree of change in the structural information. This type of method has low computational complexity , easy to implement, so it develops rapidly

Judging from the research results at home and abroad, the objective evaluation methods of image quality based on SSIM can be divided into two categories: the first category proposes new evaluation factors; the second category weights different regions of the image. , all of which do not involve the weight adjustment between evaluation factors in the SSIM model, which will inevitably affect the accuracy of SSIM in evaluating distorted images of different distortion types

Images