Pan-sharpening via the contourlet transform

The wavelet transform has been a popular choice for the spatial transformation in the pan-sharpening process. However, the wavelet transform do not represent the directional information efficiently. On the other hand, the contourlet transform, which also has a property of multiresolution decomposition similar to the wavelet, is known to provide efficient directional information and is also useful in capturing intrinsic geometrical structures of the objects. This property of contourlet transformation is very useful for images that contain geometric features. Principal component analysis (PCA) is generally used for the spectral transformation. In this paper, an alternative algorithm based on the merger of PCA-contourlet transform for pan-sharpening is presented. The efficiency of this method is tested by performing pan-sharpening of the high resolution (IKONOS and Quickbird) and the medium resolution (LandSat7 ETM+) datasets. The resulting pan-sharpened images are evaluated in terms of known global validation indexes. These indexes reveal that this method provides better fusion results than the PCA-wavelet approach.