Fusion of high-resolution satellite images
MetadataShow full item record
Two advanced image fusion techniques, a fast Fourier transform (FFT)-enhanced intensity-hue-saturation (IHS) transform method and a wavelet-enhanced IHS transform method, are developed for fusing commercial high-resolution satellite images. Instead of using a total replacement of the intensity component as in the traditional IHS transform method, the FFT-enhanced IHS transform method uses a partial replacement based on FFT filtering. The wavelet-enhanced IHS transform method combines the traditional IHS transform and a discrete wavelet transform. It uses a partial replacement based on a wavelet-based technique, where only the high frequency part of the panchromatic image is fused to the multispectral image. Ikonos and QuickBird image data are used to assess these two enhanced IHS methods. Visual and statistical comparisons of the fused images demonstrate that the enhanced IHS methods produce better results than the traditional IHS transform and principal components analysis (PCA) methods in preserving the spectral characteristics of the input multispectral image while inheriting the spatial details of the panchromatic data. This research also assesses how the resolution ratio, defined as the ratio between the spatial resolution of the high-resolution panchromatic image and that of the low-resolution multispectral image, affects the quality of the fused image. It has been shown that a resolution ratio of 1:10 or better (e.g., 1:5) is needed for successful multisensor image fusion provided the panchromatic image is not resampled to a coarser resolution. Generally, the quality of the fused image decreases as the resolution ratio decreases (e.g., from 1:10 to 1:30). However, even with a resolution ratio as small as 1:30, the fused image is still better than the original multispectral image alone. On the other hand, due to the synthetic pixels generated in resampling, more artifacts are introduced in the fused image as the resolution ratio decreases. In some cases, when the resolution ratio is small, e.g., 1:30, one might need to adjust the spatial resolution of the fused image to obtain better spectral quality by resampling the high spatial resolution panchromatic image to a slightly lower resolution before fusing it with the multispectral image.