slim@tut.cis.ohio-state.edu (Scott Whitman) (10/20/88)
A while back, someone posted a note as to why ray tracing algorithms are discussed in this newsgroup more than scanline (or other) display algorithms. Although scanline algorithms do not usually exhibit the nice illumination models that ray tracing allows,this is not an inherent problem in the algorithm, just in the implementation itself. In addition, scanline algorithms are typically an order of magnitude or even two orders of magnitude faster than raytracing algorithms. They are somewhat harder to code but the learning experience is better and the response time more gratifying. In fact, the amount of work that it normally takes to make a raytracing algorithm tolerable (handling intersections properly,using techniques to speed up the process) is about the same as the normal workload in a good scanline algorithm. As a tip, the scanline z-buffer is fairly easy to implement as a first approach and anti-aliasing can be added to it later at minimal expense (coding and time-wise). As a matter of fact, random access or what are sometimes called list priority algorithms are even easier to code but require memory (z-buffer) or a frame buffer available directly (painter's algorithm). There are various sources for simple versions of these renderers available. For instance, check the Siggraph '86 Image Rendering Tricks Course notes. Estimate time of implementation of a bare bones algorithm: 1-2 months depending on time invested - it can be done in a week or two if you are willing to spend night and day on it, but the debugging time can quickly add to that. -- Scott Whitman, Graduate Research Assistant in Computer Graphics slim@tut.cis.ohio-state.edu or ...!cbosgd!cis!tut!slim The Ohio State University; 2036 Neil Ave. Columbus OH USA 43210-1277