Analysis of Blind Image Quality Index

Abstract

Image quality index is the measure for estimating the level of degradation present in an image. Measurement of such index is challenging in the absence of reference image. Blind image quality assessment refers to evaluating the quality of an image without the need of any reference image. The quality of an image can be considered as the contrast, sharpness, brightness and other features extracted from that particular image. Other features like Discrete Cosine Transform (DCT), Wavelet Transform and Gabor filtering can also be used to extract the quality of an image. Different algorithms are developed by researchers to solve the quality evaluation problem. These algorithms are not tested on a common platform. The algorithms that are analyzed in this thesis are Blind Image Quality Index (BIQI), Distortion Identification-based Image Verity and INtegrity Evaluation (DIIVINE), BLind Image Integrity Notator using DCT Statistics (BLIINDS) & Visual Codebook. Laboratory for Image & Video Engineering (LIVE) database which is a standard database is used to analyze the mentioned algorithms. Spearman and Pearson correlation coefficients are used for validating the algorithms. Recently Visual Codebook algorithm was proposed by Peng Ye and David Doermann. The existing Visual Codebook algorithm is optimized with respect to the number of clusters used in K-Means clustering part of algorithm. Effect of variation in patch size on the performance of algorithm is studied in this thesis and an optimum value of patch size is proposed.