No-reference image quality evaluation method based on fully convolutional neural network

Abstract

The invention discloses a no-reference image quality evaluation method based on a fully convolutional neural network. A full reference image quality evaluation method is adopted to obtain the objective and real quality map of each distorted image in a training set as supervision to train the normalized image of all distorted images in the training set, and an optimal fully convolutional neural network regression training model is obtained; the normalized image of the distorted image to be evaluated is input into the optimal fully convolutional neural network regression training model; the objective quality evaluation prediction quality map of the distorted image to be evaluated is predicted; and the saliency map of the distorted image to be evaluated is used for carrying out weighted pooling on the objective quality evaluation prediction quality map so as to obtain an objective quality evaluation prediction value. Since various features, including full reference features, saliency features and the like, of the distorted image are combined and the features can accurately describe the distorted image, dependency between the objective evaluation result and subjective perception is effectively improved.

Description

technical field [0001] The invention relates to an image quality evaluation method, in particular to a no-reference image quality evaluation method based on a fully convolutional neural network. Background technique [0002] With the rapid development of image processing, machine learning and computer vision, image quality evaluation has become a research field that has attracted more and more attention in this trend, because it is an important technology that can be used in practical applications Accurately assess the quality of images. In the process of image acquisition, transmission, compression, storage, and display, there are often different degrees of distortion, such as image blur, video terminal image distortion, and image quality in the system are not up to standard. Therefore, an effective image quality evaluation mechanism should be established Very important. [0003] In general, image quality assessment can be roughly divided into two different categories: su...

AI Technical Summary

Problems solved by technology

Objective image quality evaluation methods are mainly divided into three evaluation methods, namely, full-reference image quality evaluation method, semi-reference image quality evaluation method and no-reference image quality evaluation method. The corresponding original image cannot be obtained in the application, so the research of no reference image quality evaluation method is more practical

[0004] The existing general-purpose no-reference image quality assessment methods are mainly aimed at the research of special images (for example, people, landscapes, etc.), and various types of images (including buildings, landscapes, animals, people, food, Vehicles, boats, airplanes, etc.), there are few evaluation methods for training a universal model. It is difficult to train an optimal model due to the mixture of multiple types of images (different features), so using multiple types of images to train It is more challenging to perform quality assessment with general reference-free image quality assessment methods

Images