Two Stage Fusion Scheme Based on a Curvelet with Gradient Technique for Integration of SAR Image on PAN Image to Improve the Spatial Resolution.

Abstract

    The process of combining the information captured by the various image sensors to improve the perception of an image is known as the multi sensor image fusion. In this paper, curvelet with gradient technique based image fusion scheme is described to integrate synthetic aperture radar (SAR) image onto panchromatic (PAN) image. The preprocessing of the SAR is first performed using effective speckle noise removal filter. The coregistration of these two images is then performed with respect to the PAN image. Grey level co occurrence matrix (GLCM) image texture analysis is next used on SAR image to obtain the texture feature. Then the texture feature of SAR and the PAN images are decomposed into fine and coarse coefficients using the wrap fast discrete curvelet transformation (WFDCT). Further the specific features of the SAR curvelet fine coefficients  are integrated into PAN curvelet fine coefficients. The integration of images is performed by the  block based WFDCT, which  imposes the different selection rule for each block. In the final stage, the SAR image is integrated with an intermediate PAN image using the gradient fusion technique to have more information while preserving the nonadaptive sparse representation of an image. Experimental outcomes illustrated that the fusion scheme proposed in this paper has good performance compared to other fusion methods, and  potential to be applied to integrate  the synthetic aperture radar and panchromatic  images.