Digital camera + IR LED + sensor = crop around objects based on distance?

[dompel]

Hi guys,
This might be completely retarded but the idea popped into my head after seeing some of the xbox kinect stuff. I posted this over at laserpointerforums but they recommended I asked this to a digital camera community so here I am.

Imagine you have a point and shoot digital camera, but built into it very close to the lense of the camera was a powerful IR LED which would bounce of all of the objects in the photo. The sensor would pick up the approximate distance and would then be able to remove the background of an portrait for example.

Here's an illustration:





Is this possible? Has it been done? If not, why?

If it is possible, would it mean that you could make this effect with video without a costly greenscreen setup? There must be some reason why it can't be done.

[TCav]

Off the top of my head, the camera would need a second image sensor that's only sensitive to IR, in order to do the work you want done. In addition, you'd have to be able to adjust the sensitivity of the system. For instance, in your example, you want it to include the subject that is 6 feet away, but skip the background that's 30 feet away. But suppose the background was only 12 feet away. Or suppose someone behind the subject had his or her hand on the subject's shoulder; the hand would be in the photo, but wouldn't you want to include the person that hand belongs to as well? And whet if the subject was leaning forward, or you were looking down on the subject. While you would certainly want to capture their face, but wouldn't you also want to capture their hands and feet?

I suspect that it's possible, but I think it's a lot easier to do it the way you did in the images you included above.

[dompel]

Originally Posted by TCav

Off the top of my head, the camera would need a second image sensor that's only sensitive to IR, in order to do the work you want done. In addition, you'd have to be able to adjust the sensitivity of the system. For instance, in your example, you want it to include the subject that is 6 feet away, but skip the background that's 30 feet away. But suppose the background was only 12 feet away. Or suppose someone behind the subject had his or her hand on the subject's shoulder; the hand would be in the photo, but wouldn't you want to include the person that hand belongs to as well? And whet if the subject was leaning forward, or you were looking down on the subject. While you would certainly want to capture their face, but wouldn't you also want to capture their hands and feet?

I suspect that it's possible, but I think it's a lot easier to do it the way you did in the images you included above.

Instead of doing it right from the camera, what if it took 2 pictures? An IR picture with estimated distance of everything in the field of view which it matches up to the regular photo taken at the same time. This way, later on you can say "ok well my subject including his perspective was only between 3-5 feet from my camera so anything further than that gets removed.". I don't see a need to dial it into a specific distance if the IR camera can just get a rough distance of everything for use later.

Body shots that include their feet seem like a limitation with this type of camera, but it would be really neat for webcams or something.

[VTphotog]

The difficulty I see with this technique would be the question of the distance infromation - how would it be received, and how recorded with the image? Maybe something like a laser barcode scanner? If it could be done, then there are a number of uses it could have, such as expanding or reducing depth of field, as well as automatic cropping and object selection.
Keep working on it.

brian

[TCav]

As I see it, the camera could emit a brief IR flash, and the time it took for the flash to return to the camera would indicate the distance to the object within the image. That info could be stored for each pixel, thereby facilitating the dynamic cropping, and could even be adjusted dynamically.

[MartinSykes]

I think it's easier to use two cameras - just like a 3D camera. From that you can pretty accurately reconstruct the 3D information and then cut the relevant pixels from either of the pictures. It's much more accurate than the ToF (Time of flight) approach. (Incidentally it used to be called LIDAR)

Reflected light also doesn't work very well round the sides of objects where it isn't reflected back to the camera or on shiny objects.