A Framework for Sample-Based Rendering with 0-Buffers

We present an innovative modeling and rendering primitive, called the 0-buffer. for sample-based graphics, such as images, volumes, and points. The 2D or 3D 0-buffer is in essence a conventional image or a volume, respectively, except that samples are not restricted to a regular grid. A sample position in the 0-buffer is recorded as an offset to the nearest grid point of a regular base grid (hence the name 0-buffer). The offset is typically quantized for compact representation and efficient rendering. The 0-buffer emancipates pixels and voxels from the regular grids and can greatly improve the modeling power of images and volumes. It is a semi-regular slmcture which lends itself to efficient construction and rendering. Jmage quality can be improved by staring more spatial information with samples and by avoiding multiple resamplings and delaying reconstruction to the final rendering stage. Using 0-buffers, more accurate multi-resolution representations can be developed for images and volumes. It can also be exploited to represent and render unstructured primitives, such as points, pMicles, curvilinear or irregular volumes. The 0-buffer is therefore a uniform representation for a vaiety of graphics primitives and supports mixing them in the same scene. We demonstrate the effectiveness of the 0-buffer with hierarchical 0-buffers, layered depth 0-buffers, and hybrid volume rendering with 0-buffers. CR Categories: 1.3.3 [Computer Graphics]: Pictureflmage Generation-Antialiasing; 1.3.6 [Computer Graphics]: Methodoloev and Techniaues4raohics data structures and data thc scene complexity. Image-based rendering methods use an image (e.g., a texture, a depth image [McMillan 1997; Popescu et al. 20001, a layered depth image (LDI) [Chang et al. 1999: Lischinski and Rappopon 1998; Shade et al. 19981) as the rendering primitive. All these images are digital, have a finite resolution, and are stored in a regular 2D gridded buffer. Similarly. 3D volumes, whether samples of the red world, simulation of a 3D phenomena, or voxelization of a geometric model [Kaufman 19871, have a finite resolution and are stored in a regular 3D gridded buffer. The regular sampline pattern and limited resolution of images and volumes result in limited position precision for sample points. This makes conventional images and volumes inefficient in representing precisely high-frequency details (e.g., edges of 3D objects) and in handling multiple samples in one grid point. Also, there i s an abundance of irregular samples in computer graphics, such as point clouds from range scanners, curvilinear and irregular grids from scientific computations, and samples from adaptive or jittered sampling for antialiasing. However, regular images and volumes have accuracy limitation in handling these irregular samples. As computer graphics primitives diversify, one challenge for sample-based rendering is how to organize and mix different sample primitives into one scene and render them in correct visibility order, especially when volume rendering is involved. Typically, geometric models, images, volumes, and point clouds represent some objects in 3D space. Therefore, one may think that a volume representation is an ideal framework to represent all these models. However, when pixels and points are projected back to 3D space, or