Resource- and content-aware, scalable stitching framework for remote sensing images

摘要

This paper proposes a new approach to solving the scalability problem in stitching and registration of satellite images. The image stitching process is mostly utilized to capture an image of an entire object, whose parts fall in separate but overlapping mosaics. Due to their high resolutions and huge sizes, the stitching of satellite mosaic images is time- and resource-consuming process. The work presented in this paper proposes a resource- and content-aware, scalable stitching technique. The main idea is based on measuring and assessing the available hardware resources of the machine, on which stitching process runs, and scaling the resolution of the images accordingly. Moreover, preprocessing is not needed, and the adjustment is performed dynamically at run time. The effectiveness of the proposed technique was assessed by partially registering overlapping satellite mosaic images of the island of Cyprus and Lake Van. The scalability tests are performed on varying numbers and sizes of the overlapping LandSat-8 satellite images.