A perceptual blind blur image quality metric

摘要

This paper turns on a perceptual blind blurred image quality assessment method developed in the wavelet domain. The proposed blur quality metric considers the association of an objective measure based on edge analysis through the wavelet transform resolutions and the Just Noticeable Blur concept (JNB). Unlike the existing objective metrics, the proposed one is able to assess the perceptual blurred image quality relying on the human vision system (HVS). The idea is to estimate the perceptual blur in the edge map through resolutions using the psychometric function. Tests on blurred images from different datasets provide high correlations against subjective scores compared to some existing methods.