Direct volume rendering of unstructured grids

• Published in: Computers & graphics Vol. 27; no. 3; pp. 387 - 406
• Main Authors: Berk, Hakan, Aykanat, Cevdet, Güdükbay, Uğur
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.06.2003
• Subjects: Cyclic meshes; Direct volume rendering; Early ray termination; Image-space coherency; Object-space coherency; Unstructured grids; Direct volume rendering; Unstructured grids; Early ray termination; Cyclic meshes; Image-space coherency; Object-space coherency
• ISSN: 0097-8493; 1873-7684
• DOI: 10.1016/S0097-8493(03)00034-7

This paper investigates three categories of algorithms for direct volume rendering of unstructured grids, which are image-space, object-space, and hybrid methods. We propose three new algorithms. Cell Projection algorithm, which falls into object-space category, is capable of rendering non-convex meshes through a simple yet efficient sorting schema that exploits both image and object space coherencies. Existing hybrid methods use object-then-image traversal order that enforces the processing of each cell. Thus, these algorithms perform redundant operations and do not support early ray termination. We propose a hybrid method, called Span-Buffer Ray Casting (SBRC), that can support early ray termination discarding redundant operations by employing image-then-object traversal order. Another hybrid method, called Koyamada-SBRC (K-SBRC), is proposed with the motivation of refining image-space and hybrid methods to extract the best features of them. This method is developed by blending SBRC approach with Koyamada's algorithm, which is an efficient image-space algorithm. All proposed algorithms are capable of handling acyclic non-convex meshes and generating images of acceptable quality. SBRC and K-SBRC algorithms have the additional capabilities of rendering cyclic meshes and supporting early ray termination. The proposed algorithms and Koyamada's algorithm are implemented and experimented in a common framework for analyzing their relative performance.