Design and Analysis of Algorithms for Obtaining Super Resolution Satellite Images

Abstract

This thesis dealt with the development and analysis of super-resolution algorithms for satellite images. The algorithms were developed to improve the resolution of a single image. The algorithms work by combining interpolation techniques with sharpening filters optimized for a particular image. The first algorithm was developed by reducing the size of the satellite image by a given factor and then enlarging it by the same factor by using interpolation. Then 2D filter is designed to maximize the peak signal to noise ratio (PSNR) between the original image and the interpolated image. The original image is then enlarged by the same factor using interpolation and the previous 2D filter is used to sharpen the resultant image. The second algorithms used 2D filters that maximized the structure similarity index measure (SSIM). Different filters were designed by using the two methods and their performances were assessed by testing them on a variety of satellite images. These filters were called sub-optimum filters because they are optimized for a particular image and then used for an enlarged one. The assessment in the improvement of the resultant images required the availability of the original high resolution images so that they can be compared to the sharpened interpolated images. Also the availability of these images allowed the design of optimum sharpening filters. It should be noted that in practice we do not have the high resolution original images and these are just used for comparison purposes. The performances of the sub-optimum filters were very close to that of the optimum filters. It was found that the second algorithm which maximizes the SSIM resulted in a slightly sharper images compared to the first one which maximizes the PSNR. Finally the resultant images were compared to images obtained by using the sparse algorithm. It was found that the new algorithms resulted with better images.

Date of Award	2013
Original language	American English
Supervisor	Hussain Al Ahmad (Supervisor)