Perceptual Quality Evaluation of Hazy Natural Images

Abstract

Haze is an intrusion element that disrupts the color fidelity and contrast of outdoor natural images affecting their perceptual quality. The differential characteristics of hazy images compared to other natural images restricts the generalization of existing Image Quality Assessment (IQA) algorithms. At the same time, no efficient IQA algorithm has been proposed in the literature for predicting the perceptual quality of naturally hazed images due to the lack of dataset. To address the lack of hazy natural images datasets, we build the IIT-JMU Hazy Image Dataset comprising of more than 1000 high-definition hazy natural images consisting of diverse categories such as landscape, forests, roads, seascapes, and cityscapes along with their subjective quality ratings. We present an analysis of existing Natural Scene Statistics (NSS) based IQA algorithms on hazy natural images. Finally, we propose a novel convolutional neural network (CNNs) based quality assessment algorithm for hazy natural images along with an IQA metric called deep Learningbased Haze Perceptual Quality Evaluator (DLHPQE). The proposed DLHPQE efficiently predicts the perceptual quality of hazy natural images without a reference. Our results demonstrate that DLHPQE outperforms existing state-of-the-art NR IQAs in terms of several IQA performance parameters such as PLCC, SRCC, KLCC, and RMSE. IEEE