Hi Don,
> - - Anyone have suggestions to altering a camera's FOV? Either by different
> lens (any considerations? specific lens suggestions?) or by some other
> trick? Ideally, is there some way to _make_ a camera's otherwise fixed
> FOV adjustable? (I'm thinking these CMOS board cameras, here)
I am still not sure of what do you want, sorry its probably me.
FOV is measured in degrees. If you only want to match FOVs you may
get a large FOV camera and "mask it", so only the right amount of degrees
pass to your display. Of course this may need some software or as you say
some optics.
Now, regarding the equation of image formation, it is as follows (the pinhole assumption):
If you have a 3D world point P with Cartesian coordinates (X,Y,Z)' its
correspondent image point is p=(x,y,-f)' given the formula: p=-(f*P/Z).
Where "f" is the focal distance from the pinhole point to the CCD, as in this
diagram:
| P
| /
C | f /
C |___o__ World
D | /
| /
|p
the _o_ is the point where the lens concentrates all rays of light (and the
centre of our cartesian frame), and because we assume the lens forces all rays
through this one point we assume it is a pinhole. and as you can see "f" here
is the distance from that pinhole to the CCD, and it is normally given in mm.
A large value of f (the CCD is far away from the pinhole) would mean that the
area the CCD is sensing is a small amount of the world, thus giving
a "telephoto" effect. And conversely, a small value of "f" would project a
large portion of the world into the CCD, meaning a "wideangle" effect.
And this formulation neglects
radial distortion which can also be modelled and corrected in software.
As I write this, and since current cheap and small CCD cameras have fixed value of "f" may be your
best choice would be to get a wide FOV camera and fix the amplification of the
image in software (a digital zoom) so you can match whatever "f" you want while
keeping an small CCD rig. The other alternative is to make an optical zoom
(that changes the "f").
I have never worked in Augmented Reality, those people should have standard methods for "calibrating" displays...
However, coming back to an issue one gentleman recently mention, if you are
not going to track the eye, you
may as well stick the camera as close as possible to the eye with an small offset. Say
just above the eyeball or to one side (without beam splitter), and then write your software that display the images from camera to the HMD in a way
you account for this translation. I am still not convinced of the need to
have an static beam splitter which will only match the optical axis of camera
and eye when the latter is exactly in a given (fixed) position.
Do something of this makes sense?, if not let me know and I will try again.
Hope this helps,
Walterio
--
Subscription/unsubscription/info requests: send e-mail with subject of
"subscribe", "unsubscribe", or "info" to 
Wear-Hard Mailing List Archive (searchable): http://wearables.blu.org
Please, *PLEASE* don't subscribe through a forward/expander/false domain
From Wear-Hard Mailing list Archive (WH)
Maintained by R. Paul McCarty
Archive created with babymail