Study on the spectral quality preservation derived from multisensor image fusion techniques between JERS-1 SAR and Landsat TM data

Abstract

The advantage of multisensor data fusion stems from the fact that the use of multiple types of sensors increases the accuracy with which a quantity can be observed or characterized. The response of radar is more a function of geometry and structure than surface reflection as occurs in the optical wavelengths. A suitable fusion method has to be chosen with respect to the used spectral characteristic of the multispectral bands and the intended application. This paper describes a comparative study of multisensor image fusion techniques in preserving spectral quality of the fused images. Image fusion techniques applied in this study are: wavelet, intensity-hue-saturation (IHS), principal component analysis (PCA), and high pass filtering (HPF). With these image fusion techniques, a higher spatial resolution JERS-1 SAR is fused with Landsat TM data. The merging process is carried out at the pixel level and the comparison of the resulting images is explained based on the measurement in preserving spectral quality of the fused images. Assessment of the spectral quality is performed by graphical and statistical methods between the original TM image and the fused images. The factors computed to qualify the fused images are: mean, standard deviation, coefficient correlation, and entropy. With a visual inspection, wavelet and PCA techniques seem to be better than the other techniques. PCA provided the greatest improvement with an average entropy of about 5.119 bits/pixel.