2D sub-optimum filters for sharpening interpolated satellite images by optimizing the structural similarity index measure

This paper deals with the design and analysis of 2D filters for improving the resolution of interpolated satellite images. The images are reduced first by a certain factor and then interpolated back to the original size. Linear phase 2D filters are designed to optimize the structural similarity index measure (SSIM). Then the satellite image is enlarged by the same factor and the 2D filter is used to sharpen the image. The performance of the new sub-optimum filters was assessed by using the peak signal to noise ratio (PSNR) and the SSIM on a variety of satellite images. It has been found that this method yields better results than optimizing the mean square error (MSE) or by using the sparse method.