PIWA-LOC—A Parallel Resampling Algorithm for Large Images on Cluster Systems

Image resampling is a computation-intensive task and can be found in many applic ations. To achieve better performance and generalization, a distributed parallel geometric correction algorithm is improved and a parallel resampling algorithm called PIWA-LOC is presented under cluster systems. In PIWA-LOC, the input image is partitioned evenly. Each computing node calculates the corresponding area in the output space for the local input subimage, and performs resampling for the output pixels in this area. A data structure is put forward to save irregular-sh aped output subimages, and a piece-wise linear approximation method is explored to get the area of the output subimages, which achieves good generalization for the algorithm. Experimental results show that the algorithm is suitable for many complex geometric transformations, and achieves good parallel performance for l arge image resampling tasks, especially under a cluster system with high network bandwidth.