Quantifying perceptual image quality by difference scaling

Image compression maps an original image into a bit stream suitable for communication over or storage in a digital medium. Many image compression schemes have tried, using signal models or visual perception models, to obtain efficient image coding (high compression rates), i.e., to maintain a high image quality. We propose to measure the evolution of the quality of vector quantized color images within different color spaces. We use the difference scaling technique to quantify this quality evolution. Thus one is able to make inter-space comparison to select the best color space in terms of better quality for a given compression rate or in terms of a higher compression rate for a given quality