Aerial image super-resolution based on deep recursive dense network for disaster area surveillance

Aerial images are often applied into disaster area surveillance. High-resolution (HR) aerial images are preferred to monitor the disaster area since they can provide abundant information. However, limited by hardware device and imaging environment, the resolution of captured aerial images may not meet the needs of practical application. Image super-resolution (SR) is an effective way to improve the resolution of captured images in a post-processing manner. Recently, convolutional neural networks (CNNs) have demonstrated great success in image SR. However, these CNN models cannot be easily applied to real-world scenarios due to requiring huge storage and computational resources. To reduce resource consumption, we need to decrease network parameters. Recursive network can effectively reduce network parameters, which motivates us to explore a more effective image SR method. In this paper, we proposed a deep recursive dense network (DRDN) to reconstruct HR aerial images. In the DRDN, the proposed recursive dense block (RDB) can fully extract abundant local features and adaptively fuse different hierarchical features of LR image for HR image reconstruction. In addition, the recursive manner of RDB in DRDN can effectively reduce the parameter of network. The experimental results on aerial images demonstrate the superiority of our proposed method.