Quality assessment on remote sensing image based on neural networks

Image quality assessment is of great significance for the designment and application of remote sensing systems. CNN based method is proposed for image quality assessment on remote sensing image in this paper. Specifically, we first introduce the convolutional neural network and deep learning method. Then a deep CNN architecture is constructed to automatically extract image features to evaluate image quality. Afterward, the information entropy threshold is used to remove the image blocks with less information content. Finally, a deep network model with two convolutional layers is used to achieve feature extraction and image quality scoring. The experimental results show that the quality score of this method has good subjective and objective consistency for multi-distortion remote sensing images and common multi-distortion images. Evaluation of distorted images does not depend on a specific database and has database independence. In addition, our proposed method is simple to achievement. (C) 2019 Published by Elsevier Inc.