From: jim@baroque.stanford.edu (James Helman) Subject: Re: TECH: goggle/tracker/tactile product reviews Date: 29 Feb 92 00:53:54 Organization: Stanford University In article <1992Feb28.163902.10887@u.washington.edu> bcsaic!chrise@cs. washington.edu (Chris Esposito) writes: 1. New Goggles - the Cyberface 2 from LEEP Optics appears to have it all over the VPL HRX - wider field of view, comparable pixels/degree, greater apparent resolution 1481 H by 84 V (although I don't think I completely understand how they derive these "equivalent linear resolution" numbers), both RGB and NTSC input, and a SUBSTANTIALLY lower price ($8K vs $50K). In principle, it's a great idea, but you have to be careful how you use them. The Cyberface 2 provides what LEEP claims is "equivalent linear resolution" by more dramatically increasing magnification as a function of angle. Thus pixels in the center appear more closely packed that at the periphery. This is almost what you want, except that ideally the region of high pixel density would track the eye's fovea. Most of the familiar headmounts have this distortion to some extent, it's just a little more extreme on the Cyberface 2. So if you want the best resolution, you need to turn your head to match your gaze direction. That's not so bad really. Well worth it given the limited resolution LCDs we currently have to work with. Great idea so far. Problem: if you want straight lines away from the center to appear straight, you need to render them as curves. One way of doing this without sacrificing performance is to use LEEP's camera optics, which match those of the headmount (this also buys you correction for chromatic aberration), and put video cameras in front of your hi-res graphics monitors. Alternately, you can predistort your geometry and then render it with standard 3D graphics hardware. This requires subdivision and elimination of T-vertices for polygons. The performance penalty for this might be unacceptable. Of course, you could always build your own 3D graphics system capable of doing the proper rendering. OK, you might say. I'll live with curvy things in the periphery. Whoops.... another problem. The Cyberface 2 also diverges the optics of the two eyes more than previous headmounts (hence that nice wide field of view). This puts the left edge of your right eye's field of view near the middle of your left eye's and vice versa. Thus looking straight ahead, you have one eye seeing highly distorted geometry and the other seeing less distorted geometry. Oh-oh. Image registration and stereo problems..... Going the LEEP CF2/camera/optics route probably costs less than an HRX. It also eliminates the need for scan converting to NTSC. But there's also the issue of whether a 1024x1280 display has enough pixels to both fill the camera's field of view and give adequate resolution at the center.... but this posting is way too long already. The bottom line is that there's no free lunch..... Jim Helman Silicon Graphics SGI: (415) 335-1151 (jimh@surreal.asd.sgi.com) FAX: (415) 591-8165 "The power of the computer is locked behind a door with no knob." -B. Laurel