Cross-spectrum error criterion as an image quality measure

Various criteria have been considered for image quality measure. Some of these involve subjective human visual evaluation, while others rely on statistical techniques. In this paper, a cross-spectrum error criterion is considered as a distortion measure for undersampled band-limited images. The effect of undersampling, aliasing on the perfect image, is demonstrated by manipulation in the frequency domain. Subjective comparison of the digital simulation results and the appropriate calculated spectral error indicate that the latter is a useful quantity for the analysis of distortion present in undersampled images. Relevance of our analysis to spatial filtering experiments and the mean square error is also discussed.