A Distributed Parallel Resampling Algorithm for Large Images

Image resampling is an important and computation-intensive task in many fields. In order to improve its efficiency, a distributed parallel resampling algorithm with good data locality is provided, in which each processor entirely localizes its computation through getting and resampling the corresponding area in output image for local sub input image. A data structure is put forward to save the information of irregular sub output image, and a method is presented to compute local output area. At last, all of the sub output images are gathered and stitched into the integrated target image. By implementing the algorithm on a cluster system, the results show that, for large images, this parallel algorithm improves the efficiency of resampling greatly.