Interactive Continuous Erasing and Clustering in 3D

Abstract

As an important visualization way, volume rendering is widely used in many fields. However, occlusion is one of the key problems that perplex traditional volume rendering. In order to see some important features in the datasets, users have to modify the Transfer Functions in a trial and error way which is time-consuming and indirect. In this paper, we provide an interactive continuous erasing for users to quickly get features that they are interested in and an interactive clustering way to view classified features. The first method map user's direct operation on the screen to 3D data space in real time, and then change the rendering results according to the modes that users make use of. Users could directly operate on the 3D rendering results on the screen, and filter any uninterested parts as they want. The second method makes use of Gaussian Mixture Model (GMM) to cluster raw data into different parts. We check the universal practicality of our methods by various datasets from different areas.