View Single Post
Posts: 14 | Thanked: 124 times | Joined on Jul 2010
#9
Hi all,

I'm one of the Stanford guys involved in the project. The external gyro is something Nokia Research got made to test gyroscope-using apps. It communicates to the phone over bluetooth with timestamped 3-axis angular velocities. Unfortunately the bluetooth communication isn't hugely reliable, even at such a short distance, and the whole app was a little flaky as a result.

We referred to the built-in accelerometer as insufficiently accurate because a) accelerometers don't detect rotations of the camera that cause blur (i.e. joerg_rw is right), only the linear accelerations (which are usually less of a problem), and b) we needed a fairly high data rate and high precision to detect the small movements that cause a several-pixel blur in the output image.

Our most recent version of this app just directly looks for sharp edges in images, and seems to work more reliably (it's included in fcamera as the best-of-8 burst mode). Oh well

- Andrew
 

The Following 24 Users Say Thank You to abadams For This Useful Post: