Combination of deep learning-based features for image quality assessment without reference

Images are often distorted by some necessary treatments (capture, compression, transmission, etc..) that can affect the perceptual quality. To evaluate the impact of those treatments, a plethora of metrics were developed in the literature. In this paper, we propose an efficient blind method to estimate the quality of 2D-images based on the selection of relevant patches through the saliency information and a Convolutional Neural Network (CNN). Saliency information is here used to focus on regions that highly impact the subjective scores. Three CNN models were individually evaluated and compared (AlexNet, VGG19 and ResNet50). A pairwise combination of deep learning-based features was then applied using different pooling strategies. Our method were compared to the state-of-the-art using two common datasets (LIVE-P2 and CSIQ). The obtained results showed the efficiency of our approach and its generalization ability.