Multimedia image quality assessment based on deep feature extraction

Measurement of visual quality is of significant importance to many image processing tasks. The target of image quality assessment (IQA) is to design effective computational models in order to automatically predict the quality of images in a perceptual consistent manner. We propose a full reference (FR) IQA metric based on deep convolutional neural networks and information-theoretic IQA framework. The previous proposed PAVIF is incorporated into the powerful convolutional network VGG19. Both the reference and distorted image are fed into the VGG19, and the output of each channels in the first 35 layers are utilized to measure the perceptual quality difference. The final objective score is obtained by averaging all the channel-wise quality scores. Experimental results on TID2013 and LIVE image database demonstrate that our proposed metric is competitive with many state-of-the-art IQA metrics.