A novel algorithm of remote sensing image fusion based on shift-invariant Shearlet transform and regional selection

The aim of multispectral (MS) and panchromatic (PAN) image fusion is to enhance the spatial quality of MS images and avoid spectral distortion. Considering the fact that the attribution of corresponding pixels and the local information of objects are very important for image fusion, a novel algorithm of remote sensing images fusion based on shift-invariant Shearlet transform (SIST) and regional selection is proposed. Firstly, the feature vectors of MS and PAN images are extracted and then partitioned them into regions by fuzzy c means (FCM). Secondly, the SIST is used to provide an efficient representation of the first principal component (EC1) of MS obtained by entropy component analysis (ECA) and PAN images. The low-frequency coefficients of MS image without any modification for the reconstruction level of fusion algorithm. A multi-strategy fusion rule of high frequency subbands based on regional similarity is proposed. At last, fused image is obtained by inverse SIST and inverse ECA transform. Visual and statistical analyses demonstrate that the fusion quality can be significantly improved and spectral distortion can be suppressed to a great extent by the proposed method.