Abstract
In this study, a wavelet-based method and some standard fusion methods were performed on IRS Panchromatic and multi-spectral images to benefit from its spatial and spectral advantages simultaneously. Selection of the best fusion parameters using a trial and error approach is often very time-consuming. So the best wavelet-based fusion parameters were determined using optimization Taguchi method and the importance degree of these parameters in wavelet-based image fusion efficiency were figured out. As a preliminary result, image fusion with Db4 as the wavelet type in which images were decomposed in level 4 with the Maximum and Mean functions as fusion methods for approximations and details is the best wavelet-based fusion method. The method has been compared with several standard fusion techniques according to spectral and spatial characteristics of participating bands. The results indicate that the wavelet based method presents a better spectral performance than the standard ones (R = 0.96), but PCA fusion method has a better spatial performance (RMSE = 1.06). This study also indicates that Taguchi method can be used accurately in order to optimize the performance of wavelet-based image fusion and determine the importance degree of parameters used in this method. Finally, a pixel-based and an object-oriented classification procedure have been employed to delineate barchan dunes using the fused images. The results show that object-oriented approach outperforms pixel-based approach in classifying barchan dunes (Kappa = 0.73 for pixel-based and Kappa = 0.87 for object-oriented). It is also demonstrated that the wavelet-based fused images are better inputs for classification processes than the others.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have