Robust low-illumination enhanced image quality evaluation method

Abstract

The invention discloses a robust low-illumination enhanced image quality evaluation method, which comprises the following steps: performing color space conversion on an input image, separating to obtain brightness and chrominance channel information of the image, and synthesizing a brightness image and a chrominance image according to the brightness and chrominance channel information; according to the image information obtained through separation, using a feature extraction network to extract brightness features, contrast features, color features, global structure features and texture detailfeatures of the reference image and the image after low illumination enhancement; calculating similarity coefficients of brightness features, contrast features, color features, global structure features and texture detail features of the reference image and the low-illumination enhanced image; and obtaining a weight parameter of an optimal image feature similarity coefficient through a method of training and optimizing a neural network, and adding and summing the similarity coefficients according to the weight parameter to obtain a final quality score of the image after low-light enhancement.

Description

technical field [0001] The invention relates to the technical field of image analysis and image quality evaluation, in particular to a robust low-light enhanced image quality evaluation method. Background technique [0002] Image quality evaluation is an important research direction in the field of image analysis. It hopes to use algorithms to simulate human perception of image quality differences, analyze image quality, and provide optimization directions for image enhancement technology. This technology plays an integral role in computer vision, image enhancement, image optimization and other fields. [0003] Generally speaking, image quality assessment methods are mainly divided into three typical methods: full-reference image quality assessment, semi-reference image quality assessment and no-reference image quality assessment. When evaluating the quality of distorted images, the full-reference quality evaluation can refer to all the information of the original image, th...

AI Technical Summary

Problems solved by technology

[0004] The current full-reference quality evaluation method performs well in image quality evaluation performance in different scenes, but its evaluation target is general images, and it lacks consideration of the differences between low-light enhanced images and general images in complex scenes

Usually, low-light image enhancement will face the problems of low brightness or overexposure, abnormal contrast, color shift, and loss of texture details. Therefore, the general image quality evaluation method has a large deviation in the quality evaluation process of low-light enhanced images.

[0005] In order to better adapt to the needs of image quality evaluation after low-light enhancement in complex scenes, and overcome the problems of the above-mentioned full-reference quality evaluation method, we need to consider the specific distortion problems after low-light image enhancement, and develop a method for low-light image An Enhanced Accurate Full-Reference Quality Assessment Method

Images