Projection-based Data Level Comparison of
Direct Volume Rendering Algorithms


Abstract:

There are so many variations of direct volume rendering (DVR) algorithms today -- some faster than others, some more accurate than others, and almost always resulting in slightly different images. Before we confuse and lose the confidence of our visualization users, we need to understand and explain the differences in the images. This paper extends the traditional image level comparison techniques of side by side comparison and difference images by presenting a strategy for data level comparison of DVR algorithms. Unlike image level comparisons, where the starting point is 2D images, the main distinction of data level comparison is the use raw data and intermediate 3D information available during the rendering process. The main challenge with this approach is finding an appropriate basis or framework for comparing a rich variety of DVR algorithms.

In this paper, we describe how different DVR algorithms can be viewed as projection-based DVR approaches, and how a set of projection-based metrics can be used to compare them. The key advantage of data level comparison over image level comparison is its potential to allow us to explain why and how differences arise from different DVR algorithms. We illustrate this point with a case study.

Paper:

A postscript version of the paper can be obtained by clicking here , or by anonymous ftp to ftp.cse.ucsc.edu then get pub/reinas/papers/proj/paper.ps.gz.

Images:


Return to AVIS home page.