Image quality assessment using a rotated Gaussian discrimination function

Abstract

We propose a more general adapted form of an image quality metric that was introduced as the Universal Image Quality Index (UQI) [1] and later refined into the Structural Similarity Index Metric (SSIM) [2]. SSIM is a well accepted metric that has been successfully adopted by image and video quality researchers and practitioners. It has many potential applications in modern video delivery systems. This full reference based metric is simple to use, computationally efficient and is viable as a useful mechanism for rate-distortion moderation in mobile networks including mobile game based video delivery. In this paper the authors propose a modification to the treatment of the SSIM parameter space which increases the potential discrimination capability and flexibility of the SSIM approach. The new metric was tested on the LIVE database [5] and the results show a notable improvement in performance when metric value is plotted against mean opinion score (MOS) as compared to SSIM plotted against MOS.