Illumination of the eye to detect the pupil: partial proof of concept

765 BME C90-0 BME Design

J. Browning, Chang, Morshedzadeh, and Huo

Febuary 15th, 1997

When I went through and tried to process for centroids some of the earlier pictures, I noted the following problems:

1) I got the blue glare from the monitor, and I couldn't process it out.

2) It was dang hard to process the iris free from the shadows on the face.

So, I tried some solutions. Having the screen down at chest level instead of just straight in front cut down the glare, but still makde detection difficult. Shining a light directly into the face both drowned out the glare and made the pupil easy to see (light just diappears into the pupil, and so it shows up really well as a black spot. compare the above image with the below image.

Above image: Without light

Above image: with light: note the easily seen pupil, and lack of blue glare.

These images are greyscaled, as would a b&w quickcam seeing in the infared range.

Image on right is image on left after processing. Note how the pupil stands out like a spot.

We know that using the quickcam for IR is pathetically simple. At it's most basic, we could just use a IR diode from a remote control in place of my flashlight, and the quickcam sans infrared sensor, and we can easily pluck a pupil from the image. No problemo.

here's the eye, illumated with straight on light, before negativing. The picture onthe right is a blink.

Here's the above two images after negativing. Note that nothing in the blink picture is as bright as was the negatived pupil on the left.

As for the details of how this thing goes IR, try here for how it looks and here for the gory dissection details. (The color quickcam has the same modification specifications, and heck, when we order the camera for mass production from connetix, we can ask them to just ditch the plastic frame.

The dissection of a quickcam:....(credits to hanno muller for the pics);

Here's the quickcam, sans IR filter, taking a picture of a remote control...cool, huh? : )

Basically, we know that the quickcam can view IR without difficulty. We know that when we do illumate the eye, we get the pupil to stand out, without difficulty. Unfortunately, since we can't mess with a quickcam (lack of funds) we won't be able to actually do the experiment, but we can say "if we did do this...." and then base our centroid process on the assumption we'll get images like we did above using a visual flashlight. I'm trying to schedule an appointment with dr. horswill on tuesday to get the lowdown on this.

For even more weirdness:

It turns out that the connectrix quick cam actually feeds it's data into a parallel port. What it's using the keyboard port for is -power-! The shunt cable pulls some of the power from the keyboard. Weird, huh?

The cable setup is shown above. The loose end (bottom right) goes to the camera.