Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!cs.utexas.edu!hellgate.utah.edu!cs.utah.edu!thomson From: thomson@cs.utah.edu (Rich Thomson) Newsgroups: comp.graphics Subject: Re: Fractals, and Philosophy of Science Message-ID: <1990Jan8.133924.25513@hellgate.utah.edu> Date: 8 Jan 90 20:39:24 GMT References: <119.256E54C5@uscacm.UUCP> <1247@becker.UUCP> <9144@cbmvax.commodore.com> <6780@lindy.Stanford.EDU> <9215@cbmvax.commodore.com> <12707@phoenix.Princeton.EDU> <1990Jan7.143923.8647@hellgate.utah.edu> <12782@phoenix.Princeton.EDU> Organization: Oasis Technologies Lines: 139 I wrote: :- In an attempt to bring this back around to graphics, think back on the :- methods used to generate most computer imagery in use today. Shading :- equations often contain lots of little parameters that are tweaked by the :- programmer -- often judged on "purely subjective criteria" in order to get :- the best looking image. In response, markv@gauss.Princeton.EDU (Mark VandeWettering) writes: > Agreed, but this is changing. Witness the work done with radiosity, > [...] There is a high element of physics involved in what they do. :- That's what I feel graphics is about -- getting the best image for what you :- want. > Well, that might be what you feel graphics is all about, but I > wouldn't say your classification is universal for all graphics users. You mean all graphics users don't want to get the "best" image for their needs? That's what I was saying, in case it wasn't clear. In some cases (for the purposes of a short animated film, perhaps) that image may be created from a fractal model with no demonstrable relation to physical processes, while in others it may come from a physically based model. The point is not that fractal models are more "realistic" than other things or that models based on experimental results are better than hacks. The point is that when you're producing an image, you're attempting to get a certain result or effect out of that image. Different users are creating different images with different tools for different needs. I would say that getting the best image for one's needs is universal for all graphics users (within parameters of budget and time, of course). > Some graphics programmers are botanists, and geologists, and > biologists. And physicists and artists and students and..... Right. That is why not every image need be generated from radiosity or ray-tracing techniques -- many times wire-frames will do quite nicely. Suppose an artist sketched that landscape into a paint program. Would you berate the artist because it wasn't realistic and he didn't take the time to accurately model the underlying physical structures? Of course not! :- If you want to completely and accurately :- model a mountain, you might as well sketch it by hand because the compute :- time necessary to model the complete developmental cycle of a terrain is :- prohibitive; > Well, sketching it by hand implies that you have the terrain in > front of you, which is certainly not a certain matter. The point I was trying to make is that there are trade-offs between accuracy and time when modelling objects of the complexity of terrains. For cinematic or theatrical purposes, going to highly accurate geophysical models for terrains aren't worth it. > you are saying that fractals are good at creating cheap effects > at a high level of visual complexity, then I agree. Well, "cheap" is a subjective term, isn't it? Graphics is full of "cheap" effects -- color cycling animation, for example. I think that you are falling into the trap of thinking every graphics user/programmer is a scientist concerned with the accuracy first. I don't think this is the case YET. Fractals are good at generating complex shapes from simple, concise descriptions (in terms of algorithms, etc.). Not everyone has the time to program extremely sophisticated simulations of terrain generation (I don't even think it's been done to the level needed to get even realistic looking terrains; they are just too complex). > If you say > that these are useful in computer graphics I agree that they are > useful but not the end of the evolution of graphics modeling. I never claimed that; in fact, I believe we are more in agreement than you think. > If you try to claim that there is some subjective comparison between > fractal models and the objects they mimic, then I might even yield. There is always a subjective comparison between a model of an object an the object itself. Science proposes models for which many individuals can arrive at the same subjective comparison, i.e. the results are repeatable. > If however, you claim that fractal models have some analytical > relationships to the objects they model, I merely ask to be given > proof. That is exactly the opposite of my whole posting! My whole point is that fractals are a useful tool in computer imagery in the same tradition as the lighting equation hacks from the early seventies, etc, etc. Sure, now we have radiosity and ray-tracing. Show me a workstation that allows me to rotate objects rendered with a radiosity technique with my knob box in real-time. Analytically and physically accruate models are compute intensive, so we still use those "graphics hacks" to do the job. No, they aren't completely realistic or true to the physics of light. But when a scientist is trying to understand the shape of a complex protein, one doesn't care about realism and physical accuracy in the shading process. They want to be able to manipulate that model freely. Now, on the other hand just because fractals are currently a useful tool with the status of a "hack" in computer imagery doesn't mean they have no bearing on reality. I feel that Chaos theory has more to show us in this area, but even without it similarities between fractal methods and real processes are found. Diffusion limited aggregation is one such area. L-systems (which I referred to in my previous post) are another and they've been around since 1968. :- not to say that knowledge from the sciences is useless in :- helping you obtain a good image, but that the technique used to get a good :- image doesn't necessarily have to come from a "scientifically accurate" :- model. > There are perhaps two aims to computer graphics: > 1. to "describe" a scene visually > 2. to "simulate" a scene visually I feel that the aim of computer graphics is to provide an image that satisfies the needs of the user. Sometimes that means hacking a scene together for a background in a commercial and sometimes that means simulating shock-wave propagation. These are means, not aims. > Mr. Thomson appears to say that the aim of computer graphics is 1. > I tend to believe that more of 1 is done, but the better aim is > to shoot for 2. Yes its complex and nasty, but hey, what else are > us researchers supposed to be paid for :-) Well, I'm flattered that you think I'm so important that you have to refer to me as mister ;-). I agree with you that more of 1 is done than 2 and more for the reasons of economics and finding the best tool for the job than just being sloppy researchers. Again, I think you are confusing graphics researchers with scientists. Scientists definately want simulation to ensure accuracy, yet simulation doesn't imply that you must simulate the physics of everything. Again, molecular models don't need to be ray-traced in order to be useful and accurate. Similarly, terrains and other object capable of being modelled by fractals don't have to be modelled by some other process because fractals haven't been proved to be fundamental to their creation in the real world. Use the best tool for the job. -- Rich Rich Thomson thomson@cs.utah.edu {bellcore,hplabs,uunet}!utah-cs!thomson More Colombians are killed by American cigarettes than Americans by Colombian cocaine