Applied Single Image Super-Resolution for Aerial Imagery Enhancement

One of the greatest challenges while dealing with UAV datasets for computer vision applications is the poor resolution of captured images. Super-resolution, the class of techniques that focus on improving the resolution of imaging systems, can be applied to UAV images to improve the performance across object detection, segmentation and surveillance applications. This paper attempts to apply a deep learning based method of single-image super resolution with the goal of enhancing the resolution of drone captured images. The super resolution model accepts the low resolution image as input and outputs the high resolution one through mapping that is represented as a deep convolutional neural network. The evaluation of the applied methodology was carried out using eight image quality metrics, and the results show promise for the future of deep super resolution enhanced aerial images.