Greyscale image deblurring would probably be the easiest to get a grip on and talk about. That's a pretty basic coding lesson, though deblurring can get complex if you care more for quality. Remember kids, blurring is not destructive! Block out info if you want to hide something. If you're interested in anything medical, ICT is pretty cool to learn about, though any math behind it is definitely way over my head. For those unaware, ICT is basically a kind of CAT scan that is better/worse at imaging certain things. PCA is really statistically oriented, so if getting the important bits from a data set sounds interesting to you, that'd be the choice.
If I was picking, I would probably go with PCA.