Computing distance fields is a fundamental requirement for many algorithms of data visualization and analysis. We have designed and implemented a new spatial data structure, named parallel distance tree, to enable highly scalable parallel distance field computing. The method is general to support various data types (including, but not limited to, polygonal objects, point/particle data, and volumetric data) and handle different distance metrics (including, but not limited to, Euclidean distance, City block distance, and Chessboard distance). For example, this library has been integrated with a large-scale combustion simulation code to enable in-situ processing and data reduction.
Chaoli Wang, Hongfeng Yu, Kwan-Liu Ma
IEEE Computer Graphics and Applications 30(1): 59-69 (2010)
Hongfeng Yu, Jinrong Xie, Kwan-Liu Ma
Submitted for Publication, February 2014
These research projects were sponsored in part by U.S. Department of Energy SciDAC Program through grants DE-FC02-06ER25777, DE-CS0005334, and DE-FC02-12ER26072 with program managers Lucy Nowell and Ceren Susut-Bennett.
For further information, please contact Kwan-Liu Ma.