Volume Rendering Transfer Function Design Based on eXtreme Gradient Boosting for Fast Visualization

Abstract

Volume visualization has been widely used to simulate and observe complex data in the fields of science, engineering, biomedicine, etc. One central topic of volume visualization is the transfer function (TF) of volume rendering. By setting TF, users design the mapping of voxels to optical properties of 3D datasets. However, the design of TF is usually a blind process. How to classify all volume data accurately and design a suitable TF fleetly is the key to improve the efficiency of volume rendering. In this paper, we propose a new TF design approach based on extreme gradient boosting (XGBoost) algorithm for fast visualization. First, the features are extracted from 3D volume data. Then we use the XGBoost model to classify the volume data and design TF. Finally, we assign the optical properties to the voxels to express and reveal the relevant features of dataset. This approach can help users to render the volume data efficiently. It has been tested and achieved satisfactory result.