Homogeneity-Based IKONOS Data Fusion by PCA-Wavelet Approaches
Achieving an optimum balance between spatial and spectral resolution of fused remotely sensed images has been one of the most important issues among the earth scientists. Data fusion has proposed as a solution for integration of spatial and spectral details of different data sets. However, the quality of resulting data set from the fusion process depends on many factors. Particularly, the degree of homogeneity of the landscape is an important parameter. Because of the spatial variation of man-made landscapes, spatial pattern of the fused data quality shows considerable variation; whereas for practical purposes, similarities of local qualities can be as important as the high overall and local qualities. For examination and evaluation of the effects of homogeneity of the multi spectral data, homogeneity index (HI), based on the calculation of variance in a fixed window size, has been developed for segmentation of the original data into three homogeneity classes. The study has performed on Tehran Urban Areas, enjoying multi spectral and pan data of IKONOS for the test. For close examination of the effects of resolution differences between the pan and multi spectral data, different multi spectral data sets with varying resolutions ranging from four to 44 meters have been produced through the simulation. Quality of the resulting PCA-Wavelet based fused data has been evaluated through the correlation analysis between the fused and original data sets. Results of the correlation tests have shown that there are significant differences between the fusion qualities of different homogeneity classes. Heterogeneous regions show lower correlations and higher sensitivities to degradation of the resolution of multi spectral data. In contrast, homogeneous regions show higher correlations and lower sensitivities to the resolution differences between the pan and multi spectral data.