Map Mosaicking with Dissimilar Projections, Spatial Resolutions, Data Types and Number of Bands

When researchers are interested in multiple geographic datasets over similar geographic areas, it is oftentimes necessary to mosaic the digital images. The process becomes increasingly difficult as the images vary in projection, spatial resolution, and data representation. We first present a framework to describe this mosaicking process in a formal setting. We resolve the issues of varying projection, spatial resolution, and data representation by means of mappings between sets. We also introduce a method within the framework to mosaic spatially localized digital maps that are otherwise highly heterogeneous. The motivation for our work comes from applications that require computing statistics of map attributes over a set of closed boundaries, e.g., counties. Given our motivation, the objective is to minimize the information lost from the mosaicking process while computing spatial statistics. We strive to meet the objective by proposing a new mosaicking method, defining an error metric and comparing our proposed method with the two mosaicking methods available in the ArcGIS commercial software package. Our obtained results for multiple mosaicking methods show that, for our data sets, the proposed method falls in between the two ArcGIS methods in terms of accuracy and computational requirements, and exceeds both methods in terms of easy usability.