Feasibility of high resolution SAR and multispectral data fusion

Abstract

Unlike PAN sharpening, the fusion of SAR with multispectral data involve use of non-overlapping spectral bands which poses certain inconsistencies viz. 1) radiometric differences due to their acquisition in entirely different spectral bands 2) geometric differences due to range and angular imaging of SAR and Optical sensors respectively. Apart from these, speckle noise and registration related factors will pose difficulties in the fusion. The geometric differences can be overcome to a certain extent by proper selection of acquisition parameters of the sensor. In this context feasibility conditions for proper fusion by selection of data is explained. In regard to radiometry this paper proposes a methodology that is based on multiresolution pyramids and spectral weighting functions. It has shown better performance in preservation of both spatial, spectral contents as well as better overall information content in the fused image. It has shown good balance in contrast between high frequency features of SAR and multispectral images. The results of the proposed method are compared with other methods like wavelet, Ehlers, PCA and IHS. The result of applying on TerraSAR-X SAR images with IRS-P6, Liss-4 multispectral images are analysed and illustrated in the paper.