OBJECT-BASED IMAGE FUSION METHOD BASED ON WAVELET AND PCA FOR REMOTE SENSING IMAGERY

In this paper, a new object-based wavelet fusion technique is presented for the fusion of multispectral (MS) and panchromatic (PAN) images to improve spatial information and preserve spectral information. The basic idea is to build a segmented label image by statistical region merging and minimum heterogeneity rule (SRMMHR) segmentation method to guide the object-based image fusion (OBIF). There are three key techniques of the OBIF method including SRMMHR segmentation, Mallat's and a trous wavelet transformation, and fusion rule based on object energy. The results demonstrate that the new OBIF methods based on wavelet and PCA are better comparing with pixel-based image fusion methods such as Mallat's, a trous, Mallat's PCA, a trous PCA.