A Multi-Scale Transform Method Based on Morphological Operators for Pansharpening

The aim of pansharpening is to fuse the low resolution multispectral (MS) image with the high resolution panchromatic (PAN) image to provide a synthesized MS image with high resolution. One of the main approaches for pansharpening is the multi-resolution analysis (MRA). It is generally successful in transform of spectral information. But, it often results in spatial distortion in the fused product. To deal with this problem, a morphological profile based multi-scale transform (MP-MST) is proposed in this paper which utilizes the good characteristics of morphological filters for reduction of spatial redundancies in the pansharpened image. More efficient approximate image and detail image are achieved from the MS and PAN images by applying the closing and opening by reconstruction operators, respectively. Different spatial structures with different sizes are extracted through considering a range of structuring elements sizes. The performance of the proposed MP-MST methods is compared to MST ones by doing experiments on three different remote sensors GeoEye, QuickBird and IKONOS. The experiments show the superior performance of MP-MST method compared to MST in terms of various qualitative assessments. The visual comparison is also investigated. The proposed MP-MST methods solve the problem of noise and redundant spatial information in the pansharpened images significantly.