Satellite imagery plays a critical role in recent popular Virtual Globe systems since it delivers spatially-related information in a direct and intuitive way. A satellite image may be very large in size due to large coverage, high resolution, or both, and therefore the construction of global pyramids, a core data structure of Virtual Globe, will be time-consuming if designed improperly. This article, based on the idea of divide-and-conquer, proposes an efficient algorithm, termed CGP, for the Construction of Global Pyramids, which builds global pyramids with only a single sequential scan of input imagery. By analyzing the space complexity of CGP, the memory-minimum pyramid level is derived, at which the memory requirement of CGP is minimized to a practical level, even for very large satellite images. This article also discusses a parallel implementation of CGP, which parallelizes the two main actions in CGP, thus further improving the pyramid construction performance. Both theoretical analysis and experimental results show that our approach outperforms other methods and, more importantly, this advantage increases considerably as the size of input imagery increases.
展开▼
