Contourlet Transform with Sparse Representation-Based Integrated Approach for Image Pansharpening

Pansharpening is an algorithmic approach to achieve high spatial resolution multispectral image. This paper proposes a new pansharpening method that is an integration of contourlet transform with sparse representation (CT-SR). CT is used to perform multiscale decomposition of input images to separate high-frequency and low-frequency subbands. High-frequency subbands are fused based on local energy calculation whereas SR using training dictionary is used to extract salient features to fuse low-frequency subbands. The synthesized multispectral image is obtained by combining fused low- and high-frequency subbands. It is observed that spatial resolution of synthesized multispectral image is improved, but at the same time, the original spectral resolution is also preserved. The detailed comparison between the resultant images of CT-SR and commonly used pansharpening techniques including wavelet transform with sparse representation are carried out at visual assessment level as well as through calculation of various performance indices. It is found that CT-SR performs is improved compared to other algorithms considered in this work.