Blind image quality assessment of multi-degraded images

In recent years one of the most important problems in blind image quality assessment is to achieving perceptual model that can predict the quality of distorted images completely blind. It means the model should perform without any learning process and by as little knowledge about their distortion as possible. Most previously methods measure the quality of an image degraded by a single degradation. Single degradation relies on a great degree of accuracy while, they aren't appropriate to be performed for a combination of two degradations. In this paper a new method is proposed which is able to evaluate the degradation of combination of blur and desaturation. Moreover, it has proven that the natural images have regular statistical characteristics and thus, the proposed method relies on color characteristics. These characteristics are measurably modified where distortion exists. Thus we extract some natural scene statistic features which are enabling the prediction of the image quality score without any training process.