Issues in the Design of a No-Reference Metric for Perceived Blur

Developing an objective metric, which automatically quantifies perceived image quality degradation induced by blur, is highly beneficial for current digital imaging systems. In many applications, these objective metrics need to be of the no-reference (NR) type, which implies that quality prediction is based on the distorted image only. Recent progress in the development of a NR blur metric is evident from many promising methods reported in the literature. However, there is still room for improvement in the design of a NR metric that reliably predicts the extent to which humans perceive blur. In this paper, we address some important issues relevant to the design as well as the application of a NR blur metric. Its purpose is not to describe a particular metric, but rather to explain current concerns and difficulties in this field, and to outline how these issues may be accounted for in the design of future metrics.