Hi Matt, That looks interesting. Looks like the same technique I tried a few years ago at the Imiloa stereo dome in Hawaii where I used 24000 virtual cameras so that each vertical strip had the stereo cameras facing that strip. This paper describes camera hardware for videoing the real world so it uses just 3 fish-eye cameras and there are artifacts because of that. The author says the artifacts are much reduced with 5 cameras. It would be a good technique to try even for virtual rendering where you can get away with thousands of cameras, because it would be a lot faster to render. You’d especially want that for interactive rendering in stereo dome. My stereo dome rendering took 15 minutes per frame with the 24000 camera approach! http://www.cgl.ucsf.edu/Outreach/technotes/dome3d/dome3d.html Tom On Apr 8, 2014, at 12:53 PM, Dougherty, Matthew T <matthewd@bcm.edu> wrote:
Hi Tom,
came across this
http://vision3d.iro.umontreal.ca/en/blog/2013/07/29/the-omnipolar-camera-a-n...
Matthew Dougherty National Center for Macromolecular Imaging Baylor College of Medicine ________________________________________ From: Tom Goddard [goddard@sonic.net] Sent: Monday, April 07, 2014 8:18 PM To: Dougherty, Matthew T Cc: chimera-dev@cgl.ucsf.edu Subject: Re: [chimera-dev] dome camera/S3D
Hi Matt,
Yes the parallax angle used for Chimera dome stereo with the stereo command is half the “convergence angle” if by convergence angle you mean the angle between the lines of sight from your two eyes to an object on the dome surface level with your eyes and midway between them. So if I was setting it for a viewer in the center of a 4 meter radius dome, with eye separation of 0.06 meters I would use an angle atan(0.03 / 4) = 0.43 degrees.
The positive angle is for the right eye negative for the left eye, because the y-axis points up and a positive rotation in the standard mathematical conventions (right-hand rule) would rotate your straight ahead view line to instead point slightly leftward. The view line is supposed to intercept an object at dome distance midway between eyes. A simple test in Chimera dome mode using “stereo dome parallax 10” verifies that this gives a view from the right as should be seen by the right eye.
It makes sense to use symmetrical values +d degrees for the right eye and -d degrees for the left eye. If instead you shifted the two parallax values so their sum is not zero, that would give an effect as if your head was facing a different direction in the dome other than straight ahead.
The Chimera focal plane setting seen in the Top View mode of the Side View dialog (menu Tools / Viewing Controls / Side View) does correspond to the dome screen position in the straight ahead position. I say in the “straight ahead position” because the focal plane is a flat plane, and of course the dome is a spherical surface. The focal plane position is important. Setting the parallax value does two things. It moves the camera right or left along the x-axis, and it rotates the camera view direction. The shift along the x-axis is by an amount so that the view line will hit the mid-point on the dome screen, ie (x camera shift) = (focal distance) * tan(parallax). The Chimera “eye separation” parameter is not used for dome parallax mode. The reason for that is because the dome parallax mode isn’t a stereo mode — it renders only one image using one camera. You have to record your whole movie with a right eye setting, and then re-record it with a left eye setting, then combine the two movies with some external software. Since it is a mono mode being used twice, it doesn’t use the “eye separation” parameter in the Camera panel (menu Tools / Viewing Controls / Camera). Instead the eye separation is implied by the parallax value and focal plane distance.
I realize this is all somewhat quirky. If we provided a real interactive dome stereo mode that for instance produced side-by-side left and right eye dome images, then we could do away with parallax and use the eye separation and focal distance parameters that are used for other Chimera stereo modes. Since the fish-eye projection generally needs to be warped before display on the dome, this interactive stereo mode would probably also need to accept a warping texture described in a file in order to be useful.
Tom
On Apr 7, 2014, at 2:10 PM, Dougherty, Matthew T wrote:
Hi Greg,
I was experimenting with the dome parallax and needed some help understanding your sw and terminology.
Reading the documentation: Left- and right-eye stereo views for dome display can be generated by specifying parallax angles of opposite signs. The views are rotated by p-angle degrees about the Y (vertical) axis through the center of the focal plane. For example, views recorded with dome parallax ±5° can be combined to give a stereo effect when viewed in the forward (−Z) direction.
When you talk about p-angle it reminds me of the angle of convergence. Is that correct?
In hollywood terminology they calibrate the word "parallax" to percentages. Zero parallax is on the screen, positive is behind the screen, and negative in front of the screen.
+5° would be the left eye, and -5° the right? Proper usage would be +n and -n; that is symmetrical around 0? as opposed to +5° and 0°, or +4 and -2?
Is the focal plane actually the dome screen? Setting the focal plane in the side view, does it actually do anything?
Matthew Dougherty National Center for Macromolecular Imaging Baylor College of Medicine
_______________________________________________ Chimera-dev mailing list Chimera-dev@cgl.ucsf.edu http://plato.cgl.ucsf.edu/mailman/listinfo/chimera-dev