From: dsawdai@guitar.engin.umich.edu (Donald James Sawdai) Subject: Re: TECH: Accommodation vs convergence in stereoscopy Date: 27 Feb 1996 19:49:13 GMT Message-ID: <4gvn7p$lls@srvr1.engin.umich.edu> Organization: University of Michigan (CAEN) From: dsawdai@guitar.engin.umich.edu (Donald James Sawdai) In article <9602242100.AA13071@athena.ulaval.ca>, Marc Bernatchez writes: > Okay, did someone followed me up to here? He he, I hope so. To > conclude, this discussion has some crucial importance because it will > permits us to better analyze the way HMDs helmets should be designed > on their optical side. More precisely, it will permits to answer the > question regarding what should be the ideal virtual image plane > distance. I.e., the distance at which the HMD's pictures are > shown. Some HMD uses the infinity image plane approach, other uses a > setting of a few meters from the observer's eyes. Okay, I'll toss in my (totally unqualified) $0.02 worth... I think that for practical application in HMDs (or random dot stereograms or any other stereoscopic media), you need to consider that the brain (and hence visual perception and optical motor control) is adaptive. Most people have a fixed accomodation/convergence relationship "programmed" into their brain from many years of looking at the real world, where you focus and converge on the same point in 3-space. Even this isn't infallible, though. Try looking at a bunch of random ropes, let's say, outside (away from other objects), about 20 feet away from you. You can focus on the ropes, but converging on a specific rope is sometimes difficult (or try branches and leaves high in trees, but the overlapping objects makes this case easier to see). Anyway, my point is that the brain can learn to converge at a wide range of angles on a fixed accomdation in an HMD. This is one reason that it is easier to see stereoscopic images once you "get used to looking in 3D," and experience workers in 3D can fuse the images much quicker. In an HMD, you are stuck with a fixed focal plane. So accomodation will give you no clue for the depth of objects in the image. But for untrained (per above) users, they (IMHO) will attempt to converge on the focal plane by reflex. So I would say that the focal plane should be near the Z-median of your objects of interest. -Don Sawdai dsawdai@engin.umich.edu