A novel framework for seamless mosaic of Cartosat-1 DEM scenes

Abstract

Digital elevation model (DEM) is used as a source of elevation data in wide range of applications. The applications involving larger area than an individual DEM scene often need a mosaic of DEM scenes. A simple stitching exhibits seams along the borders of DEM scenes in the mosaic output. The presence of seam lines varies with the number of scenes in an overlapping region and their overlap extents. The conventional feathering-based blend method, by virtue of the method itself, considers only two input scenes in an instance of the mosaic process. This method requires a suitable ordering of the input DEM scenes, especially in the overlapping regions containing more than two scenes. In this paper, we propose a framework that automatically handles the multiple overlapping regions and creates a seamless mosaic of Cartosat-1 DEM scenes for a given region of interest. We use a tree data structure to arrange input DEM scenes hierarchically based on their metadata. A traversal algorithm is devised to establish an order of precedence among the DEM scenes. This determines a unique input sequence of DEM scenes during mosaic process. In addition, two variants of feathering-based blend method are formulated to handle all possible cases of the overlapping regions. The experiments have been conducted on Cartosat-1 DEM scenes over 1°×1° region of interest. We analyze and compare our proposed methodology with average method and feathering-based blend method. We demonstrate that our framework exhibits a smooth transition of elevation values in the overlapping regions of the DEM mosaic output.