Weibull distribution based model behavior in color invariantspace for blind image quality evaluation

Abstract

In this paper, we analyze Weibull distribution based fitting using phase congruency components obtained via color invariance approach. Since such model is found suitable for image quality evaluations, we test the effects on different distorted images. Blind or no-reference image quality assessment enables testing without having available subjective ratings. Thus, it is possible to consider additional type of distortions usually neglected in available datasets through data augmentation artificial intelligence tools. The obtained results show that Weibull based modelling using color invariance domain components may be suitable for further experimental analysis. Moreover, data augmentation not used in the training phase, demonstrate the importance of development modern quality evaluator that can be applied even for modified data obtained as a result of augmentation.