Survey of full-reference image quality metrics

Abstract

This paper aims to give a survey of full-reference image quality metrics, including metrics specifically designed to evaluate image quality, but also metrics for image difference, image fidelity, and more. These metrics have in common that they try to predict the perceived difference between an original image and a modified version of it, this modification can typically be compression, halftoning and blurring. They output one numerical value and/or an image difference map. More than 100 image quality metrics have been reviewed and categorized, and short descriptions and analysis of all metrics and their relationships are given. This should prove valuable to researchers in various fields, to find the most appropriate metric for their application, and to give a better understanding of the state of the art of the field of image quality metrics.