Edge Directed Radial Basis Function based Interpolation Towards PCA based PAN-sharpening

Abstract

ABSTRACT In this research, we proposed a new PAN-sharpening algorithm by using edge-directed Gaussian Radial Basis Function (RBF) and Principal Component Analysis (PCA). Up-scaling is a procedure to upgrade the spectral data of a Low Resolution (LR) Multi Spectral (MS) imagery. Non-linear up-scaling methods generate fewer artefacts in the interpolated image as they are edge-adaptive in nature. Adaptive interpolation techniques consider different features of an image during up-scaling. These techniques are proven efficient compared with non-adaptive methods. However, they may generate distorted output in fine-textured areas and can have lower luminance information. We consider the Laplacian of Gaussian (LoG) or Marr-Hildreth edge detection operator to preserve missing structure and luminance information. It generates smooth edges by minimizing artefacts in the interpolated image. We have considered multi-source and multi-sensor datasets obtained from LISS-IV, CATROSAT-I, and Quickbird to perform all experiments. We have carried out a broad quantitative and subjective assessment and comparison of the proposed algorithm with the standard and cutting-edge methods.