People have been working on that for years and have published algorithms here: http://iris.usc.edu/Vision-Notes/bibliography/contentspeople.html#Face%20Recognition,%20Detection,%20Tracking,%20Gesture%20Recognition,%20Fingerprints,%20Biometrics
But for a class assignment like this appears to be you can't be expected to write something super robust like people who did Ph.D. dissertations on it. So what I would do is to simply threshold every image in the database and extract the largest blob (demo attached below in blue). Then calculate it's area and perimeter. Every image in the database will have a unique area. So, given an "unknown" image from the database, you can identify it by calculating its area and comparing it to the known areas of all the images:
theIndex = find(area == databaseAreas);
You only need to consider perimeter if there is more than one image with the same area. If it's an image not from the database, you can include the perimeter and build a feature vector [area, perimeter] and get the distance of your feature vector from the vectors of all the database images, to find out which database image seems to be "closest" to your test image. If it's not working, add other features, like solidity or majorAxisLength, until it works.
