Hybrid Visualization of Sparse Point-Based Data Using GPGPU

Direct point-based rendering is a popular method in scientific visualization, since the number of point-based datasets increased dramatically in the past few years. At the same time, rendering of point primitives is becoming less efficient as the data size increases. Point splatting, volume-based rendering, or is surface extraction are well-known approaches that can be utilized. Unfortunately, high visual accuracy is often sacrificed. Furthermore, unstructured sparse point-based data is more difficult to visualize, since no surface geometry and topology can be implicitly defined. This paper introduces a novel hybrid visualization method that utilizes point-based and volume-based rendering of sparse point-based data. The visualization is done entirely with a custom rendering pipeline by using GPGPU, which provides accelerated rendering. The method achieves real-time visualization, while retaining high visual accuracy, as shown on cosmological dark matter SPH-based dataset.