Enhancement of satellite images using deep learning algorithms

Today, the information obtained by satellite images is used inmany areas such as civil defense operations, mineral exploration, agricultural production and geographical information systems. The accuracy and usability of the information obtained from the images are related to the resolution of the image and the detail levels of the components. For this reason, many image enhancement algorithms have been developed in literature to increase the resolution and quality of the images. Image enhancement algorithms are the process of increasing the level of components and resolution of the image by processing low resolution image, and today it is more known as super resolution. Deep learning, one of the most popular areas of study recently, also shows high performance in super resolution problems. For this reason, deep learning based super resolution algorithms have been the most researched area recently. In this thesis, high resolution satellite images were obtained by using deep learning based super resolution algorithms developed to increase the quality and resolution of satellite images obtained using existing imaging equipment. In addition, a new deep learning based algorithm has been proposed for super resolution. The proposed algorithm was compared with image quality measurement metrics referenced with other algorithms and the results obtained were evaluated.