Anti-Parallax Calibration

Post Reply
Frederf
Posts: 3
Joined: Sun May 16, 2010 11:15 pm

Anti-Parallax Calibration

Post by Frederf » Sat Nov 13, 2010 11:08 pm

As I was double-sided taping my TIR5 onto my monitor frame left side at approximately TC Pro level it occurred to me... why am I doing this?

Well I knew why, because I was getting some pretty bad parallax when leaning forward with it in the advertised top dead center position with the TCP hanging off the side of my headset. What I mean is that I know the software could be doing this all for me while leaving the camera alone.

I imagined a situation where you hit "calibrate" on the TIR software and it would run you through a few runs through each axis to develop a set of coordinate axes that was tuned to your actual head positions as opposed to apparent camera dot movements.

You lean forward as directly to and away from the screen as you can a few times. The software takes this data and constructs the necessary transforms (simple linear or sine regressions of the raw axes with plenty of averaging over 2-5 runs). Now, what was once a coupled motion due to parallax better correlates real head motion with the virtual head motion.

I realize TrueView and the new v5 software that allow coupled curves tries to deal with this, but it's a little more work than what's really necessary for the user.

With a "learned" axis real head movement in one direction that appears to the software as forward, to the right, and with a bit of a twist suddenly becomes movement along a single virtual axis.

"Camera Image" --(Image to Clip Transform)-->
"Clip Motion" --(Anti-Parallax Transform)-->
"Head Motion" --(Game Profile Transform)-->
"Game Motion"

With this extra transform in place the sensor could be placed anywhere with respect to the user's head/clip within reason.

Post Reply