Variographic Analysis of Multi-source Remotely Sensed Imageries by Wavelet-based Approaches

In the times of wide uses of commercialized high resolution satellite imageries, urban remote sensing is regarded as one of the important application fields. Currently, the main subjects and issues in urban remote sensing focus on the various attempts and approaches for these types of image data sets, distinguished from conventional image processing in remote sensing. Three types of different spatial resolution images were taken into account in this study: IKONOS 1 m, KOMPSAT 6.6 m and Landsat 15 m panchromatic imagery. Using these, we tried to investigate the pattern of spatial distribution of images using the variogram analysis. We used both theoretical variogram modelling, which is typically used, and wavelet analysis, which is composed of wavelet decomposition for approximation and detail components. Wavelet transform is profitable for analyzing individual image. However, it is difficult to compare every wavelet decomposition data to grasp similarity of a certain image with respect to other image. So, we used variogram modelling coefficient. For the variogram modelling, we tried to compound model which composed with logarithm model and nugget effect model. Results of the variography analysis reveal the spatial characteristics of each image. And Euclidean distance method using the variogram modelling coefficient was useful to search similarity characteristics of applied images. With this result, variographic analysis based on continuity computation can be considered as the effective method for characterization of urban features.