On efficient opacity correction for over-sampled volume ray casting visualization
Timothy S. Newman, Jong Kwan Lee · 2008
This dissertation focuses on investigating the correction of an artifact in oversampled volume ray casting visualizations. Specifically, new correction techniques for over-composited opacities are presented for two sampling scenarios commonly used in volume ray casting visualizations. While the only existing opacity correction technique in the literature follows a dataset homogeneity assumption, the new opacity correction techniques described in this dissertation are fast, generalized cell-by-cell approaches which introduce new opacity correction factors to avoid assuming dataset homogeneity. In addition, efficient approaches using commodity hardware to accelerate the opacity correction techniques are explored. One class of approaches exploits a programmable graphics processing unit (GPU) in performing the arithmetic operations necessary to correct the opacities. Another class exploits cluster computing environments lacking programmable GPUs. In particular, the cluster-based computation exploits multi-processing and multi-threading that utilize data parallelism. A third class is a hybrid approach which employs multi-threading using a dual core processor and a GPU. For comparison, the efficiency approaches are also extended to the existing opacity correction technique. Experimental results and analyses using volumetric datasets to measure the effectiveness of the approaches developed in the dissertation are also presented.