Blind image quality evaluation method based on combining gradient signal and Laplacian of Gaussian (LOG) signal

Abstract

The present invention discloses a blind image quality evaluation method based on combining a gradient signal and a Laplacian of Gaussian (LOG) signal. The method comprises: firstly extracting a gradient modulus signal and a LOG signal of an image; and normalizing the two signals by using a joint adaptive normalization process. Based on the two characteristic representations, three blind image quality evaluation methods which can predict image quality degradation are proposed. In each of the three methods, model training is performed on a public image subjective quality evaluation database by using a method that supports vector machine regression. A large number of experimental results prove that the three algorithms proposed by the invention all have good predict performance on the blind image quality evaluation. Compared with other existing algorithms, the model 3 method has the best quality evaluation performance and optimal database stability. Furthermore, the three algorithms are high in operation efficiency.

Description

【Technical field】 [0001] The invention belongs to the field of image processing, and relates to a blind image quality evaluation method, in particular to a blind image quality evaluation method combining a gradient modulus and a Laplacian of Gaussian (Laplacian of Gaussian, LOG) signal. 【Background technique】 [0002] Visual information is the most important way for human beings to obtain information about the surrounding environment. With the rapid development of digital image technology and network, image information has increasingly become an important way of information recording, transmission and storage. However, image quality degradation will inevitably be introduced in various stages of image acquisition, compression, transmission, and storage. With the popularity of various low-end image acquisition devices (such as smart phones), the degradation of image quality is even more inevitable. An algorithm that can quantitatively evaluate the subjective quality (or degr...

AI Technical Summary

Problems solved by technology

Due to the lack of original reference information, this method is also the most difficult, and it is not yet possible to obtain good prediction results

Images