efficient generation of transfer function for direct volume rendering

Abstract

The process of transfer function generation is a crucial step in the direct volume rendering pipeline. The transfer function is responsible for setting visual properties such as color and transparency for each voxel in the volumetric data set. Therefore, it is considered the most important part of the direct volumetric rendering process, as it is a complex process and takes a long time. That is why the transfer function is the process that potentially determines the efficiency of the volume rendering process as a whole. In this thesis, we proposed an automatic design of a transfer function based on the similarity of features of volumetric data, where the features of volumetric data are extracted through the similarity between iso-surfaces. Then, we classified these features through the affinity propagation algorithm to automatically extract the optimal number of clusters that best reflect these features. The efficiency of the proposed system is demonstrated by its accuracy in exploring the features of volumetric data and its ability to classify them automatically without the need of user intervention or pre-define the number of clusters