| Digitized virtual human is of great significance for scientific research and promising for various applications. 3D reconstruction is one of the key techniques of the research for Virtual Chinese Human (VCH). Volume rendering has the advantage of displaying the interior details of data sets, but it costs greater computation than other algorithms such as surface rendering. As we know, VCH data sets are tremendous and difficult to be reconstructed on personal computer (PC). Therefore, parallel volume rendering algorithm for VCH is studied in this thesis.A parallel ray-casting volume rendering algorithm for Virtual Chinese Female No.1 (VCH-F1) datasets is presented and implemented based on MPI (Message-Passing Interface).The Ray-Casting method is discussed in detail. 3D geometric transform is applied to image space instead of volume space to reduce the time of the calculation. A rotation method rounding variable axis in 3D space is used to achieve more convenient interaction. In order to accelerate rendering, tri-linear interpolation and threshold classification are introduced. Phong model is used in shading calculation for making the image vivid. Front-to-back composition is utilized to reduce the times of sampling in Ray-Casting. The results of the experiment, which are done on the head and feet datasets of Virtual Chinese Female No.1 (VCH-F1), show that vivid 3D images can be attained by this algorithm.In this thesis, a parallel algorithm is discussed with its strategy based on the Ray-Casting method. The data division method is based on image space, where full volume data are copied to each of the slaves to accelerate rendering. The load balance is implemented via taskpools, and the size of distribution tasks is determined by pre- experiment.Finally, the parallel performance of the parallel algorithm is analyzed. Results of the experiments show that this algorithm can provide good scalabilities of the number of hosts and the task size. |
PDF Full Text Request