An Image Quality Evaluation Method Based on Unsupervised Learning

Abstract

The invention discloses an image quality evaluation method based on unsupervised learning. In the training stage, according to the HOG feature statistical histogram of each sub-block and the mean value of the amplitude of all pixel points under different center frequencies and different direction factors, the obtained Image feature vector, and obtain the image quality vector according to the objective evaluation prediction value of each sub-block; then according to the image feature vector and image quality vector of each sub-block, construct image feature dictionary table and image quality dictionary table respectively through unsupervised learning method; In the test phase, the sparse coefficient matrix of each sub-block in the test image is obtained through optimization according to the image feature dictionary table, and the objective evaluation value of the image quality is obtained through the sparse coefficient matrix and the image quality dictionary table, which is relatively consistent with the subjective evaluation value. Good consistency, and no need to calculate the image feature dictionary table and the image quality dictionary table, which reduces the computational complexity, and at the same time does not need to predict the subjective evaluation value of each evaluation image.

Description

technical field [0001] The invention relates to an image quality evaluation method, in particular to an image quality evaluation method based on non-supervised learning. Background technique [0002] With the rapid development of image coding and display technologies, the research on image quality evaluation has become a very important link. The goal of the research on the objective evaluation method of image quality is to keep consistent with the subjective evaluation results as much as possible, so as to get rid of the time-consuming and boring subjective evaluation method of image quality, which can automatically evaluate the image quality by computer. According to the reference and dependence on the original image, the objective image quality evaluation methods can be divided into three categories: full reference (Full Reference, FR) image quality evaluation method, partial reference (Reduced Reference, RR) image quality evaluation method and no reference ( No Reference...

AI Technical Summary

Problems solved by technology

At present, the existing method is to predict the evaluation model through machine learning, but its computational complexity is high, and the training model needs to predict the subjective evaluation value of each evaluation image, which is not suitable for practical applications and has certain limitations.

Images