Slab-based Intermixing for Multi-Object Rendering of Heterogeneous Datasets

Abstract

The visualization of multiple 3D objects has been increasingly required for recent applications in many industrial, biomedical, and scientific fields. Due to the heterogeneity in data representation or data configuration, it is difficult to efficiently render multiple objects in high-level performance. In this dissertation, we propose a multi-object rendering method based on a novel intermixing model for high performance visualization of multiple volumetric and polygonal objects. To represent multiple layers of polygonal surfaces and volumetric fuzzy surfaces, we introduce zSlab model which is defined with visibility color, slab depth and finite thickness per pixel. zSlab is based on radiosity spreading out the viewing direction, which is a special ray-segment used for an intermixing unit. As the proposed zSlab model is designed to treat multiple polygonal and volumetric geometries which may be unordered, we apply the order-independent-transparency (OIT) concept to the construction of zSlabs. With the proposed zSlab-based intermixing method, this enables image-level intermixing for a variety of object combinations

traditional polygon rendering for many transparent surfaces, hybrid rendering for polygon and volume, and multi-volume rendering. First, we present how the zSlab is applied to volumetric fuzzy surface as well as infinitely thin surface of polygonal geometry with the proposed virtual zSlab concept. And, we introduce a novel z-thickness buffer that stores the zSlabs as an array, which is used for the rendering input and output. We introduce two versions of in-slab visibility interpolation methods and verify which model is well suit for the proposed intermixing algorithm. Finally, based on zSlab model and the in-slab visibility interpolation, we propose an efficient slab-based visibility intermixing algorithm so that the entire intermixing leads to a high performance rendering result (acceptable image result and fast rendering speed). Experimental results demonstrate that the proposed method delivers more effective multi-object rendering in terms of taking advantages of the image-level intermixing especially for a rendering scene that includes at least one volumetric object, providing acceptable image quality compared to the image based on the conventional on-the-fly intermixing. And the proposed intermixing method is able to resolve traditional intermixing artifacts such as aliasing intersection and z-fighting plane occurred in intersecting or overlapping surface region. Moreover, our experimental results manifest a potential of the proposed method that it can deliver a substantial aid in scientific visualization of entire context including hidden (or, inside) structures in the multi-object scene.