Skip to Main Content
Reliable methods for measuring the perceived image quality are needed to evaluate visual artifacts brought about by digital compression algorithms such as JPEG. This paper presents an objective quality-assessment method based on a circular back-propagation (CBP) neural structure: the network is trained to predict quality ratings, as scored by human assessors, from numerical features that characterize images. As such, the method aims at reproducing perceived image quality, rather than at defining a comprehensive model of the human visual system. The neural model allows one to decouple the task of feature selection from the mapping of these features into a quality score. Experimental results on a public database of test images confirm the effectiveness of the approach.