A CNN-based Quality Model for Image Interpolation

Abstract

Image interpolation techniques have aroused wide attention, which is dedicated to improving the resolution of image and providing a better visual perception. However, how to evaluate the perceptual quality of interpolated images is still an ongoing problem. In this paper, a no-reference method built on Convolutional Neural Network (CNN) is proposed for interpolated image quality assessment. To enhance the performance, we incorporate attention modules with the proposed network to facilitate feature extraction and quality prediction. Experimental results show that the proposed method outperforms related IQA metrics in perceptual quality evaluation of image interpolation.