A color image quality assessment method based on multi-channel deep convolutional neural network

Abstract

The invention provides a color image quality evaluation method based on a multi-path deep convolutional neural network, and the method comprises the steps: (1) carrying out the multi-scale transformation and color space transformation processing of a color image, and outputting a plurality of different component images; (2) designing and improving a single-path deep convolutional network structure; (3) training and optimizing a single-path deep convolutional network; (4) carrying out feature extraction and multi-dimensional feature collaborative fusion on the plurality of component images by asingle-path deep convolutional network module; (5) carrying out feature dimension reduction processing on the multi-dimensional output feature vectors; and (6) carrying out function mapping on the subjective opinion score and the dimensionality reduction feature by using a nonlinear regression method, establishing a color image quality prediction model, and carrying out color image quality evaluation. By constructing the multi-channel deep convolutional neural network model, the quality evaluation method of the natural image is realized, the prediction precision of the color image quality isimproved, and the method can be applied to the fields of dynamic monitoring and quality adjustment of images, videos and display services and the like.

Description

Technical field[0001]The invention belongs to the field of natural image processing, and relates to a color image quality evaluation method based on a multi-path deep convolutional neural network.Background technique[0002]Objective image quality measurement is an important issue in vision science. It plays a major role and influence in many image engineering applications such as image processing algorithm measurement and parameter optimization, image dynamic monitoring and adjustment. Non-reference image quality evaluation is an important research content in the field of objective image quality measurement. The objective quality evaluation of natural images can be achieved without prior knowledge of image quality distortion. At present, most natural image quality evaluation methods are designed for grayscale images. In the process of implementing quality evaluation on color images, the following two methods are usually used. The first is to convert the color image into a grayscale i...

AI Technical Summary

Problems solved by technology

Most of the objective and natural image quality evaluation methods proposed in the existing domestic and foreign related literature are designed for grayscale images, ignoring the research on the correlation between image color information and quality distortion, and the quality evaluation accuracy of color images is low

Images