No Reference Image Quality Assessment by Designing Fuzzy Relational Classifier Using MOS Weight Matrix

Assessing quality of distorted/decompressed images without reference to the original image is a challenging task because extracted features are often inexact and there exist complex relation between features and visual quality of images. The paper aims at assessing quality of distorted/decompressed images without any reference to the original image by developing a fuzzy relational classifier. Here impreciseness in feature space of training dataset is tackled using fuzzy clustering method. As a next step, logical relation between the structure of data and the soft class labels is established using fuzzy mean opinion score (MOS) weight matrix. Quality of a new image is assessed in terms of degree of membership value of the input pattern corresponding to given classes applying fuzzy relational operator. Finally, a crisp decision is obtained after defuzzification of the membership value.