Google Goggles does something similar to Amazon Remembers. It uses OCR if the text can be identified, and they want to use it with a similar search for images from Google images. I think they generate some kind of hash for the image with a function that, if the images are similar, the images are similar to.
My best guess is to try starting with character recognition and doing a text search for the name of your card. This means that your user needs to make a very clear image, possibly even at a specific position. But for the first application, that would be great. When someone played magic, I would buy a tool to trade and catalog my cards.