Image quality prediction in a multidimensional perceptual space

Abstract

We demonstrate that the relationship between images degraded by noise and blur on the one hand and judgements on noisiness, perceived blur and overall quality by subjects on the other hand can be characterized in a multi-dimensional perceptual space. In this perceptual space, the images are represented by points, and the strengths of their perceptual attributes are modeled by the projections of these image positions onto attribute axes. In analogy with the perceptual space, we introduce a psychometric space in which the positions of the images are determined by objective measures on the images. We show that this psychometric space can also be used to predict perceptual image quality and its underlying attributes noisiness and perceived blur.