Viewpoint Selection Using PSO Algorithms for Volume Rendering

In volume visualization, locating optimal viewpoint is important for the improvement of data understanding, especially for non-interactive visualization of large dataset. This paper presents a novel approach for view selection in volume rendering by utilizing particle swarm optimizers (PSO). The search for the optimal viewpoint is reformulated as a global optimization problem. The view directions are encoded as the particles in the PSO, and the optimal viewpoint is generated by the PSO iterations. Although there is no consensus about what a good viewpoint means, the quality of a viewpoint is intuitively related to how much information its corresponding view gives us about a scene. We use viewpoint entropy to define the informative view. To improve the efficiency, either the computation for viewpoint entropy or the rendering for every viewpoint is accelerated based on graphic processing unit (GPU). This method remarkably eliminates the reluctant viewpoint evaluations, thus improves the application´s performance.