An Innovative Pansharpening Method Based on MRF Strategy

An innovative pansharpening method is proposed to selectively extract more useful information from the original images to produce a new image with higher resolution information. The standard PCA is employed as a decorrelation tool to separate the spectral and spatial information in MS images. In order to reduce the spectral distortion of fused image, we decompose the first principal component (PC1) of multispectral (MS) images and panchromatic (PAN) images using nonsubsample shearlet transform (NSST) to achieve effective detailed information; a novel energy function, including the inter- and intrainformation between subbands, has been established to take full account of the local dissimilarity between MS and PAN images, and the reasonable coefficients are selectively chosen based on Markov random field (MRF). It is found that the simulated image by the new method is more close to the real image and more clear and with more detailed information compared with other popular methods reported recently, which means that our new method can effectively improve the efficiency and quality during the fusion image process.