IEEE Asilomar Conference on Signals, Systems and Computers, November 2002

Blind Quality Assessment of JPEG2000 Compressed Images

H. R. Sheikh¹, Zhou Wang¹, Lawrence Cormack² and Alan C. Bovik¹

¹Laboratory for Image and Video Engineering (LIVE), Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX 78712-1084

²Department of Psychology, The University of Texas at Austin, Austin, TX 78712-1084

PDF File (63K)

Abstract: Measurement of image quality is crucial for many image-processing algorithms, such as acquisition, compression, restoration, enhancement and reproduction. Traditionally, image quality assessment algorithms have focused on measuring image fidelity, where quality is measured as fidelity with respect to a ‘reference’ or ‘perfect’ image. The field of blind quality assessment has been largely unexplored. In this paper we present an algorithm for blindly determining the quality of JPEG2000 compressed images. As far as we are aware, this is the first attempt of its kind to design an algorithm that can evaluate image quality for JPEG2000 compressed images without the reference image. Our algorithm assigns quality scores that are in good agreement with human evaluations. Our algorithm utilizes a statistical model for wavelet coefficients and computes features that exploit the fact that quantization produces more zero coefficients than expected for natural images. The algorithm is trained and tested on data obtained from human observers, and performs close to the limit on useful prediction imposed by the variability between human observers.