Psychovisual correlations with multifractal measures for wavelet and wavelet packet progressive image transmission

Abstract

In an effort to develop a quantitative measure for image quality, this paper looks for psychovisual correlations of image quality to multifractal measures. Image metrics such as peak signal-to-noise ratio are not well suited as perceptual indicators and other techniques are primarily limited to just noticeable differences which limit their use in general. It is desirable to have image quality metrics for progressive image transmission that are more general and can evaluate images produced during a progressive image transmission, from the worst image reconstruction step all the way to the final perfect image. We focus on how the Renyi (1955) dimension spectrum changes as more wavelet and wavelet packet coefficients are included in a progressive image transmission. Mean opinion score (MOS) results serve as an additional basis for analyzing how the progressive image transmission affect the Renyi dimension spectrum. While this paper does not give a final answer to what the best metric would be for image quality in a progressive image transmission, it does attempt to link perceptual quality with multifractal measures.