Quality Assessment Method for Image Redirection Based on Statistical Similarity and Bidirectional Saliency Fidelity

Abstract

The invention discloses an image redirection quality evaluation method based on statistic similarity and bidirectional significance fidelity. The image redirection quality evaluation method is characterized in comprising the following steps of: 1) obtaining the natural scene statistical characteristics of the log-Gabor domains of an original image and a redirected image, and obtaining a difference value of the natural scene statistical characteristics of the log-Gabor domains of the original image and the redirected image as a natural scene difference value statistical characteristic; 2) obtaining a forward significance information missing value and a backward significance information missing value; 3) obtaining the significance structure fidelity value of the image; 4) forming redirection quality evaluation characteristics by the natural scene difference value statistical characteristic, the forward significance information missing value, the backward significance information missing value and the significance structure fidelity value; and 5) utilizing a support vector regression model to train and predict the redirection quality evaluation characteristics so as to obtain a quality evaluation model. By use of the image redirection quality evaluation method, the change of the natural scene statistical characteristics is considered, in addition, the significance fidelity of the image is measured from two directions, and the subjective visual quality of the redirected image can be more accurately predicted.

Description

technical field [0001] The invention relates to the fields of video signal processing and image and video quality evaluation, in particular to an image redirection quality evaluation method based on statistical similarity and bidirectional significance fidelity. Background technique [0002] With the popularity of mobile terminals, more and more device screens with different resolutions are produced in daily life, for example, mobile phones, tablets and computers all have different resolutions. The popularity of mobile terminals has also brought about a problem: how to present the same image with the best display effect on screens with different resolutions. This requires adaptive processing of image and video content according to different screen resolutions. In order to solve the problem that the same image does not match on different resolution screens, many researchers have proposed a variety of image redirection algorithms, such as cropping, scaling, warping and conten...

AI Technical Summary

Problems solved by technology

The popularity of mobile terminals has also brought about a problem: how to present the same image with the best display effect on screens with different resolutions

At present, the research on the quality evaluation algorithm after image redirection is still in its infancy. The prediction accuracy of many image redirection quality evaluation algorithms is not satisfactory, and the performance is often affected by different image distortions.

Published an article "Sift flow: Dense correspondence across scenes and its applications" in the top international journal IEEE Transactions on Pattern Analysis and Machine Intelligence in 2011. This article proposed a SIFT-flow algorithm to measure the structural similarity between two images. But the algorithm ignores information loss distortion in redirected images

In 2000, an article "The earth mover's distance as a metric for image retrieval" was published in the top international journal International Journal of Computer Vision. This article proposed a minimum loss measurement algorithm EMD to measure the mutual transformation between two distributions. The distance of the feature distribution between two images, but also ignores the information loss distortion in the redirected image.

Images