GPGPU Based Non-photorealistic Rendering of Volume Data

Abstract

Nowadays, non-photorealistic volume rendering has become a useful technique in medicine and scientific visualization. One of these rendering techniques is silhouette extraction of iso-surfaces. This paper proposes three methods of extracting silhouettes from relatively large datasets very fast (in some cases, even in real time), using the GPGPU technology. These methods are suitable for different types of datasets, applications and hardware characteristics. The first method extracts the iso-surface and then computes its silhouette. The second one extracts only the silhouette and computes the visibility of each contour vertex using an algorithm inspired by ray casting. The third method uses a CUDA rasterizer in order to render iso-surfaces and silhouettes from large datasets.