Blind Image Quality Estimation Using Deep Neural Networks Explicit Image Position

Abstract

Blind image quality assessment (BIQA) is very challenging problem due to the unavailability of a reference image. State of-the-art BIQA methods usually learn to find the image quality by regression from human subjective scores of the training samples. These methods uses to large number of human scored images for training, and lack an explicit explanation of the image quality is affected by image local features. We propose a family of image quality assessment (IQA) models based on natural scene statistics (NSS) is predict the subjective quality of a distorted image without reference to corresponding distortion less image and without any training results on human opinion scores of distorted images. Different from most deep neural networks (DNN), We take biologically inspired generalized divisive normalization (GDN) instead of rectified linear unit (ReLU) as the activation function. We empirically demonstrate that GDN is effective at reducing model parameters while achieving similar quality prediction results. The proposed model is extensively find the large scale benchmark databases to deliver high performance with state-of-the-art BIQA models as well as with some well-known full reference image quality assessment models. The proposed algorithm uses natural scene statistics in spatial domain for generating wake by distribution statistical model to extract quality aware features. The features are fed to an SVM regression model to predict quality score of input image without any information about the distortions type or reference image.