akcs.shaulp@tronsbox.UUCP (Shaul Peleg) (03/14/90)
Can anyone explain to me or lead me to an explaination of exactly, algorithmically, how to render voxels? How is it done? What are the requirements? Is ray-tracing the only way? Thanks, Shaul Peleg
markv@gauss.Princeton.EDU (Mark VandeWettering) (03/14/90)
In article <25fd9fbe:1119comp.graphics@tronsbox.UUCP> akcs.shaulp@tronsbox.UUCP (Shaul Peleg) writes: > Can anyone explain to me or lead me to an explaination of exactly, >algorithmically, how to render voxels? > How is it done? What are the requirements? Is ray-tracing the only way? Volume rendering is still a technique that is in its infancy, but some techniques have proven to be useful. Defn: A voxel is a volume element. When one speaks of volumetric rendering, typically a voxel is a small region of space which is filled with a material of varying optical properties, usually density or emission. The combination of many voxels builds up a volume based model. The value of a pixel in the final image is a function of all the voxels that pierce a view ray. (I use raytracing lingo here a bit, but it by no means indicates that raytracing is the only way to render these). When you raytrace volume models, typically you sample the ray at various points along the ray and then use this to reconstruct a continuous density curve in the ray parameter. This you can then use to calculate the final value of the pixel. Levoy's paper uses basically this approach. Scanline methods are also possible. You can render individual voxels at a time, carefully scan converting them individually, and using compositing operators in the manner of Tom Duff to build up the final image in back to front (or front to back) order. If your data is sampled on a regular grid, you can take the image processing or "Pixar" approach (my nomenclature). You can use three dimensional analogues of the traditional image processing techniques of filtering and edge detection to construct three dimensional "images" which are then rendered directly. Try the following papers to get started: Drebin, Carpenter, Hanrahan, "Volume Rendering", Proceedings Siggraph 88 (There are two other papers on volume rendering in this proceedings) Levoy, "Display of Surfaces From Volume Data", IEEE CG&A Porter and Duff, "Compositing Digital Images", Proceedings Siggraph 84 These should give you enough to get started. Mark VandeWettering
akcs.shaulp@tronsbox.UUCP (Shaul Peleg) (03/16/90)
Thanks very much. Can you tell direct me to any sampe source, if any exists? Thanks again! -=SP=-