3D Image Quality Assessment Based on Local Entropy and Disparity Map

Abstract

The aim of full reference objective quality assessment methods is to deduce a perceptual model that can delicately estimates the quality of a distortion image in a manner likely to human opinion. In this article, we have studied the efficacy of using robust features extracted from 3D stereoscopic image, to present a full-reference (FR) objective image quality assessment (IQA) model. The essential concept of the proposed objective quality assessment model is based on exploiting the statistical and spatial features of 3D stereoscopic image. More specifically, the extracted features are local entropy values and disparity map information. The proposed quality assessment model namely ED-QA was tested on LIVE 3D stereoscopic image database and compared to other full-reference (FR) objective quality assessment methods. The performance evaluation of ED-QA model was achieved by applying it over symmetrical and asymmetrical distorted images with three types of distortions (JPEG, Gaussian Blur and Fast Fading). Based on the experimental results, ED-QA model demonstrates an efficient and accurate quality measurement of 3D stereoscopic images under all the distortion types utilized in this work.