Fast Discrete Curvelet Decomposition with Gradient Fusion Based Technique for Pansharpening of Multispectral Image 
Leelavathi H P1, J Prakash2
1Leelavathi H P*, Associate Professor, Department of ECE, Vivekananda Institute of Technology, Bangalore.
2Dr. J Prakash, Prof and Head, Department of ISE, Bangalore Institute of Technology, Bangalore, Karnataka State, India.
Manuscript received on June 01, 2020. | Revised Manuscript received on June 08, 2020. | Manuscript published on June 30, 2020. | PP: 892-901 | Volume-9 Issue-5, June 2020. | Retrieval Number: E9888069520/2020©BEIESP | DOI: 10.35940/ijeat.E9888.069520
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: : The Process of improving the local details of multispectral image using the information captured from different image sensors is called pansharpening. In this paper, the information captured from three different satellite sensors such as synthetic aperture radar (SAR), multispectral (MS) and panchromatic (PAN) of the same scene is combined effectively using the proposed a novel approach. This approach consists of three stages of fusion scheme, where block based wrap fast discrete curvelet decomposition technique is used in the first stage, which results an intermediate PAN image that contain the almost complete spatial details of SAR image and nonadaptive sparse details of the SAR and PAN images. Next the gradient fusion based technique is used to combine the SAR and an intermediate PAN image which retains more information in the PAN-SAR (PS) fused image. Further pansharpening of the MS image is accomplished by using the hybrid technique which is based on injecting the primary and secondary high frequency components into the original MS image. In this algorithm, the high frequency details are extracted by taking difference between the PS and synthesized intensity image, then the extracted high frequency details are modulated by local adaptive and post processing fusion parameter. The experimental outputs reveal that the proposed algorithm results better performance than other techniques by providing more spatial information in multispectral image while preserving the spectral details and has the potential characterize urban areas in a fused image in a better way.
Keywords: Curvelet Transform, Gradient Fusion, Hybrid Pansharpening, Multi-sensor Image Fusion, Multispectral, Panchromatic, Spatial Resolution, Synthetic Aperture Radar.