Full reference image quality evaluation method based on combination of spatial-domain and frequency-domain analysis

Abstract

The invention discloses a full reference image quality evaluation method based on the combination of spatial-domain and frequency-domain analysis. The method comprises firstly performing color space conversion on reference and distortion images in a database; secondly extracting reference image and distortion image spatial-domain gradients and frequency-domain phase features to calculate a globalmaximum structural feature similarity; calculating frequency-domain texture and spatial frequency feature similarity, and spatial-domain color feature similarity, and forming a 9-D feature vector in combination with the global maximum structural feature similarity; establishing a regression model according to random forest RF to fuse the feature vector and an subjective MOS value, and performing training; and finally extracting the 9-D feature vector of an image to be tested and inputting the 9-D feature vector into the trained regression model to achieve high-precision image quality prediction so as to achieve objective image quality evaluation. The method realizes high-precision objective evaluation of the full reference image quality, and can maintain high consistency with human visualcharacteristics.

Description

technical field [0001] The invention belongs to the technical field of image processing and image quality evaluation methods, and relates to a full-reference image quality evaluation method based on spatial domain combined with frequency domain analysis. Background technique [0002] With the rapid development of multimedia, image processing and communication technologies, digital images, as one of the most intuitive and effective information carriers, convey important visual signals. At the same time, the popularity of related image acquisition and processing equipment, such as digital cameras, computers and smart phones, vividly and vividly describe objectively existing things for users. [0003] Users are always eager for high-quality images. However, in the process of image acquisition, storage, compression, transmission, and restoration, various unavoidable factors will cause image distortion and quality degradation, such as : During the shooting process, problems such...

AI Technical Summary

Problems solved by technology

[0006] At present, most image quality evaluation methods based on feature similarity calculations quantify image quality in the spatial or frequency domain, but these methods only use a single spatial or frequency domain feature information, ignoring the The characteristics of processing information in the spatial and frequency domains, resulting in low accuracy of evaluation results

Images