Jump to Main Content
A novel iterative PCA–based pansharpening method
- Ghadjati, Mohamed, Moussaoui, Abdelkrim, Boukharouba, Abdelhak
- Remote sensing letters 2019 v.10 no.3 pp. 264-273
- color, multispectral imagery, panchromatic imagery, principal component analysis, spatial data, wavelet
- Image pansharpening methods are usually grouped into two main classes: the spectral methods and the spatial methods. For the first class, the multispectral image undergoes a spectral transformation and then one of the resultant components is totally substituted with the panchromatic image, hence leading to a considerable color distortion compared with the second class. In the literature, this issue is addressed by integrating the wavelet transform to the spectral methods in order to transfer only the spatial details of the panchromatic image. Furthermore, the spatial information quantity transferred during the fusion is usually defined by the resolution ratio between the multispectral and panchromatic images, and this is, however, not necessarily the optimal quantity providing the best images. Therefore, a simple iterative Principal Component Analysis (PCA) based method is proposed in this letter, to continuously transfer the spatial information from the panchromatic to the multispectral image until the best fused image is obtained. The spatial distortion Dₛ of the Quality with No Reference (QNR) index is used as a stopping criterion. The experiments applied on the Worldview–3 images show that the suggested method presents the best visual and numerical results comparatively to the PCA and the Additive Wavelet Principal Component (AWPC) methods.