eugene@nas.nasa.gov (Eugene N. Miya) (01/31/91)
In article <28420@cs.yale.edu> musgrave-forest@cs.yale.edu (F. Ken Musgrave) writes: > Holly Rushmeier writes note about new European journal removed >>I'd like to see the discussion move forward in time >>from talking about an image that was after all published 7 yrs ago. >>Also, in case no one has mentioned it, the 1984 picture and the >>side by side conference rooms Greg Ward did appear in Foley, >>van Dam , Feiner and Hughes. >>And finally, as far as Turing tests go, there is only one paper >>that I know of where anybody did direct side by side tests of >>real and synthetic images (our obscure TOG article from '86). If >>people on the net know of similar experiments since then, unpublished Execellent reference to Greg. > Hear hear! - that kind of side-by-side comparison, particularly in the >kind of carefully controlled environment of the radiosity experiments, >approaches the real practice of Science. I'd like to move on, too. But historical notes: 1980: SIGGRAPH had Nelson Max (LLNL) showing some very impressive models of metal pieces deforming. While these are not models of radiosity, they have significance combining information in several fields materials science as well other other aspects of physics. Check earlier SIGGRAPH sidesets. They were side by side comparisons of REAL metal pieces and rendered models. Then before a film like "LA the Movie" there was "Economars Tours" [a non-Max, LLNL SIGGRAPH film]. Just check the old slide sets. What I wonder is: if we are still roughly in the same place 10 years from now (perhaps we generate images faster, but marginal gains in 'quality') and have people looking back in the same way (forgetting earlier work). --e. nobuo miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov {uunet,mailrus,other gateways}!ames!eugene AMERICA: CHANGE IT OR LOSE IT.
rthomson@dsd.es.com (Richard Thomson) (01/31/91)
> In article <28420@cs.yale.edu> musgrave-forest@cs.yale.edu > (F. Ken Musgrave) writes: > >>And finally, as far as Turing tests go, there is only one paper > >>that I know of where anybody did direct side by side tests of > >>real and synthetic images (our obscure TOG article from '86). If > >>people on the net know of similar experiments since then, unpublished From lectures Michael Cohen has given here at Utah, they did side-by-side comparison of radiosity generated images and the real-world objects at Cornell. -- Rich
stam@dgp.toronto.edu (Jos Stam) (02/01/91)
Eugene N. Miya writes: > >1980: SIGGRAPH had Nelson Max (LLNL) showing some very impressive >models of metal pieces deforming. While these are not models of >radiosity, they have significance combining information in several fields >materials science as well other other aspects of physics. >Check earlier SIGGRAPH sidesets. They were side by side comparisons >of REAL metal pieces and rendered models. Then before a film like "LA >the Movie" there was "Economars Tours" [a non-Max, LLNL SIGGRAPH film]. >Just check the old slide sets. > >What I wonder is: if we are still roughly in the same place 10 years from now >(perhaps we generate images faster, but marginal gains in 'quality') and >have people looking back in the same way (forgetting earlier work). > Is the synthesis of images undistinguishable from 'reality' possible? It seems that the amount of information to be processed is just too large. Furthermore, most current rendering methods entirely rely on physics. In human perception there is simply more involved then physics. In my opinion, some part of the rendering should be controlled by a user. Finally, is the perfect simulation of 'reality' necessary? Alternate rendering methods are sometimes more adequate in certain situations, e.g. in scientific visualization. But I agree that it is very exciting to produce images which look like the real stuff... Jos
deadman@garnet.berkeley.edu (Ben Haller) (02/01/91)
In article <1991Jan31.120714.22611@jarvis.csri.toronto.edu> stam@dgp.toronto.edu (Jos Stam) writes: >Is the synthesis of images undistinguishable from 'reality' possible? It seems >that the amount of information to be processed is just too large. Furthermore, >most current rendering methods entirely rely on physics. In human perception >there is simply more involved then physics. In my opinion, some part of the >rendering should be controlled by a user. Finally, is the perfect simulation >of 'reality' necessary? Alternate rendering methods are sometimes more >adequate in certain situations, e.g. in scientific visualization. > What is involved in human perception besides physics??!? Certainly conciousness has a role to play, and consciousness might not fall under some definitions of "physics", but if the photons coming into a human eye are the same for a real scene and for a computer-generated copy of that real scene, how could the human possibly tell them apart? What else do you think is involved? Perfect simulation of reality is necessary if, for example, you want to be able to go to a movie and *not* be able to tell your non-graphics-savvy friends which parts of the movie were artificially generated and which parts weren't. -Ben Haller (deadman@garnet.berkeley.edu)
uad1077@dircon.co.uk (Ian Kemmish) (02/04/91)
In article <1991Feb1.082556.8553@agate.berkeley.edu> deadman@garnet.berkeley.edu (Ben Haller) writes: > Perfect simulation of reality is necessary if, for example, you want to >be able to go to a movie and *not* be able to tell your non-graphics-savvy >friends which parts of the movie were artificially generated and which >parts weren't. > >-Ben Haller (deadman@garnet.berkeley.edu) Unless you are already the sort of purist who leaps up in the middle of the first reel and says `Hey, they painted the dirt on the scenery wrong', I would say that perfection is *not* necessary for film applications <-: and I am totally ignoring the limited dynamic range and colour gamut that film has relative to `reality' that no-one has so far mentioned:->. Very little is what it seems in cinema. Pixar know this: checkout various comments in various papers about `building scenes for CG is a lot like ubilding movie sets: you don't model what you can't see' and the immortal `reality == 80,000,000 polygons'. Note how they have published a lot fewer papers in the years since they achieved their commercial aim of providing reasonably seamless integreation between live action and synthetic imagery (models as well as CG). (Personally, I think the paper clip picture deserves to be far more famous than `1984' anyway...) -- Ian D. Kemmish Tel. +44 767 601 361 18 Durham Close uad1077@dircon.UUCP Biggleswade ukc!dircon!uad1077 Beds SG18 8HZ United Kingdom uad1077@dircon.co.uk