Visual Saliency Based Blind Image Quality Assessment via Convolutional Neural Network

Abstract

Image quality assessment (IQA), as one of the fundamental techniques in image processing, is widely used in many computer vision and image processing applications. In this paper, we propose a novel visual saliency based blind IQA model, which combines the property of human visual system (HVS) with features extracted by a deep convolutional neural network (CNN). The proposed model is totally data-driven thus using no hand-crafted features. Instead of feeding the model with patches selected randomly from images, we introduce a salient object detection algorithm to calculate regions of interest which are acted as training data. Experimental results on the LIVE and CSIQ database demonstrate that our approach outperforms the state-of-art methods compared.