Is there an Apple Patent for this ?

[drsox]

I'm wondering if anyone has seen an Apple patent for fixing the obvious problem with FaceTime/Skype etc.
It's obvious what the problem is - any video conversation involves not looking at the camera but at the screen, so the person at the other end is always not looking at you.
I guess to fix it there needs to be more than one camera with some form of position interpolation so that both ends look as if they are looking at each other.

[aziatiklover]

Lol que?

[KnightWRX]

So what if there is a patent ? No need to have a patent to fix an obvious issue.

But to answer your question, yes there is, and it's old :

http://arstechnica.com/apple/2009/01...tebook-screen/

[drsox]

Originally Posted by KnightWRX

So what if there is a patent ? No need to have a patent to fix an obvious issue.

But to answer your question, yes there is, and it's old :

http://arstechnica.com/apple/2009/01...tebook-screen/

Great. I wonder when it will appear in a useable product ?
Is there a current fix for this problem that I have missed ?

[ChristianJapan]

I think the only fix right now is: look into the camera and don't look the the screen.

Of course I could imagine a real-time image analysis and change the black of your eyes in the video stream.

But that would require still some solid CPU power to make it nice looking and smart. As the head in front of the camera move in 3D while the image on the screen is a 2D projection it's not very simple.

Maybe on a Mac; not so soon on a iDevice

[drsox]

Originally Posted by ChristianJapan

I think the only fix right now is: look into the camera and don't look the the screen.

Sort of defeats the object if that is required.

I guess it is 2 generations of GPU away for now.

[ChristianJapan]

One other chance is to increase the distance between camera and screen ... Like with real VC system ... But that will also not work with iPad. No zoom in optical system.
So I don't see a good solution for you right now.

But I agree with you: when I call my wife via FaceTime she always looks absence ... A bit disturbing but got used to it.

[thekev]

Originally Posted by drsox

There needs to be more than one camera with some form of position interpolation so that both ends look as if they are looking at each other.

I'm not sure interpolation is the right word there. You would have to correct for parallax issues and digitally correct for distortion. Interpolation is basically guessing the answers from known data points. It looks like the patent Knightwrx listed is just trying to capture a wider range then crop in to center the face after correcting for parallax. That isn't meant to be too nitpicky.

[drsox]

Originally Posted by thekev

I'm not sure interpolation is the right word there. You would have to correct for parallax issues and digitally correct for distortion. Interpolation is basically guessing the answers from known data points. It looks like the patent Knightwrx listed is just trying to capture a wider range then crop in to center the face after correcting for parallax. That isn't meant to be too nitpicky.

Yes, I agree. Interpolation is my shorthand word for a whole bunch of image correction needed to ensure a realistic image is created. It should be straightforward (albeit complex) to correct for parallax. The algorithm needs to do the same as the lens in tilt/shift architectural cameras.

I guess the big problem is knowing where the actual face is located relative to the centre of the screen and to the camera(s). With tilt/shift lenses it's the photographer who is doing the positioning - the GPU/CPU will need to do the same based on sensors.