Satellite Image Super Resolution using Gradient-Prior

Abstract

Super-resolution (SR) has been used in the realm of remote sensing to improve the resolution of an image and get more detailed spatial information than the original image captured by the sensor on the acquisition device. Several SR methods with different approaches, only focusing on sharpening the edges and forgetting non-edge areas. One of the SR methods that utilize prior gradients, can produce high resolution (HR) images in a short time and produce sharp images for non-homogeneous areas. But for areas that tend to be homogeneous, a lot of noise appears. This problem will affect the remote sensing process due to the amount of noise that arises. This paper offers to use dynamic weighting on the gradient prior that will reduce the noise on the homogeneous area, while still able to maintains to produce the sharp edges in non-homogeneous areas. An experimental comparison is conducted on both homogeneous and non-homogeneous area using the previous method and the proposed method.