Image quality blind evaluation method

Abstract

The invention provides an image quality blind evaluation method. The image quality blind evaluation method comprises the steps of 1), extracting a characteristic of a quality-reduced image block in a training image, and estimating an offset between the characteristic of the image block and the characteristic of a non-quality-reduced image block; 2), analyzing different types of quality reduction in a probability latent semantic analysis method, and mapping the different types of quality reduction to different theme distribution characteristics, wherein the different types of quality reduction comprise single quality reduction and hybrid quality reduction; 3), establishing a relation between an image theme distribution characteristic and the image quality based on the image training set according to a machine learning method, thereby forming a quality blind evaluation model for a hybrid quality reduction image; and 4), evaluating the quality of the quality reduction image outside the training set by means of the quality blind evaluation model. The image quality blind evaluation method has advantages of improving accuracy in no-parameter quality evaluation, setting a problem of evaluating the hybrid quality reduction image in engineering, and realizing high suitability of comprehensive evaluation for image acquisition, compression and transmission performance in a multimedia system on the condition that an original image cannot be acquired.

Description

technical field [0001] The invention relates to the technical field of image quality evaluation, in particular to a blind image quality evaluation method. Background technique [0002] Image subjective quality evaluation is a research hotspot and has a very wide range of applications. The acquisition, transmission and decoding process of image / video will introduce certain distortion and reduce the subjective quality of the image. The computer-aided evaluation of image quality can help to find the design defects of the multimedia system, guide users to adjust the system, and then ensure that users get satisfactory image quality. There is a gap between the traditional image quality evaluation method based on signal-to-noise ratio and people's subjective perception of the image, which cannot reflect the image quality well. In recent years, the research on image quality evaluation is mainly divided into two categories: with reference and without reference. [0003] The parame...

AI Technical Summary

Problems solved by technology

[0006] In view of the shortcomings of the prior art described above, the purpose of the present invention is to provide a blind image quality evaluation method, which is used to solve the problem that t...

Images