No-reference Perception Based Image Quality Evaluation Analysis using Approximate Entropy

Abstract

Due to extensive relevance across many disciplines, interest of no-reference image quality evaluation has been increased. The main goal is to assess the visual quality of an image using an objective metric that should be highly consistent with the subjective scores given by viewers. Well- known naturalness and perception based metrics include patch level distortion estimation and may show specific effects when comparing to high difference mean opinion scores. In this paper such effects are demonstrated, as well the possibility of using approximate entropy to overcome such manifestations. The obtained results show that approximate entropy technique can be used as an estimator in order to additionally distinguish image information related to subjective index.