A Parallel Approach for Volumetric Reconstruction

N. Sakamoto, J. Nonaka, and K. Koyamada (Japan)


Volumetric reconstruction, Volume intersection, Voxel col oring and PC cluster system


In this paper, we describe the development of a volumetric reconstruction system consisting of a set of cameras and a PC cluster. Volume intersection method is one of the most popular method for reconstructing 3D volume of an object from multi-viewpoint images. In addition, in this method the voxels of the object being reconstructed can be processed independently making parallelization easy. Some parallel volume intersection methods which run on PC clusters have been proposed so far, enabling real-time processing. In this paper, we propose a parallel voxel color ing method to be used with the parallel volume intersection method in order to make possible the displaying of the re constructed model without sacrificing the performance. By unifying these two parallel methods we developed a vol umetric reconstruction system which enables image cap turing, volume reconstruction and displaying at interactive frame rate. In our experiment, by using a hardware configu ration of five cameras and six-node PC cluster we obtained an average throughput of 2.63 FPS when using a volume resolution of 963 and voxel size of 10mm. This clarify the effectiveness of our proposed parallel voxel coloring method.

Important Links:

Go Back