Multispectral and panchromatic images fusion using the Markov-random-field-based FCM

This article proposes a multispectral (MS) and panchromatic (PAN) images fusion approach exploiting local spatial information by using fuzzy c-means clustering algorithm based on the Markov random field (MRFFCM). The standard principal component analysis (PCA) technique is first employed to transform the MS images into principal component spaces to extract the first principal component (PC1). Then, we decompose the PAN image using the à trous wavelet transform to get the high frequency detailed information and the approximation of the PAN image. In the process, the local relationship is employed through MRFFCM between the two to produce a fused PC1 by choosing the saliency and significant coefficients. The fused MS image is generated after the detailed information has been incorporated with the fused PC1 and finally the inverse PCA is implemented. Experimental results demonstrate that the proposed approach improves the quality of fused images both qualitatively and quantitatively.