Image quality and noise evaluation

Abstract

The definition of a 'good image' is subjective and depends on the requirements of a given application Gonzalez, R. C., Woods, R. E. (1992). For example, image quality is highly connected to the process of image sampling and data compression. Noise can be generated and added in during both processes, plus it can also be generated if further processing is imposed on the image such as brightness enhancement or contrast stretch. The common practice is to evaluate the quality of the image visually. This is a subjective process since noise cannot be measured accurately Parker, J. R. (1997). In this paper, we propose using the coefficient of determination for the unreplicated linear functional relationship (ULFR) model, namely R/sup 2//sub F/ as a measure of the similarity between two images which in turn may be used as a definition for image quality Dolby G. R. (1976), Fuller, W. A. (1987). The derivation of R/sup 2//sub F/ is briefly reviewed. This study shows that the proposed similarity measure performs better than particular similarity measures Dietrich et al. (2002).