Unsharp Masking Based Pansharpening of High Resolution Satellite Imagery

Abstract

Pan sharpening is a pixel-level fusion technique for increasing the spatial resolution of low-resolution multi-spectral satellite imagery using high-resolution panchromatic imagery. In this work, the performance of various pansharpening algorithms in improving the 2.5m resolution of Gokturk-2 satellite imagery is compared. Eight different pansharpening algorithms are tested on seven different Gokturk-2 images and evaluated using eight different metrics which measures spectral and spatial quality of pansharpened images. We developed a preprocessing method, which is based on unsharp masking of pan band. Proposed method replace high pass filtering of Optimized HPF method instead it used sharpened image from unsharp masking of pan band. Then this information is added multispectral bands based on their standard deviation ratios. Pan band carries highest resolution spatial information while it lacks color. Unsharp masking is applied pan band to improve spatial contribution from this band. Then other pansharpening methods are applied to the image. Unsharp based preprocessing of pan bands improves sharpness of the images without degraded image quality. Unsharpening increases sharpness of pansharpened images which performs poorly persevering spatial features. Our results show that the Optimized High Pass Filter (HPF) yields the sharpest pan sharpened image, while the hyper spherical Color Space (HCS) method preserves the truest colors while preserving sharpness to a certain degree.