Welcome to the fun world of computer vision, which is basically the field you deal with image processing with AI algorithms to do recognition/tracking/segmentation etc. Without being exhaustive, the whole deal involves image acquisition, image processing, image segmentation, image recognition.
The easiest way to start is probably OpenCV. In most applications however, using OpenCV is probably overkill, as it's really not that difficult to start from scratch. Once you got past the picture loading (You probably would wanna consider BMP as it's easiest to parse) part, you can implement many basic image processing algorithms if you just had the mathematical formula. There are also other library you could probably try, stuff like torchvision, AForge (.NET) are good choices.
When it comes to the AI part, here's where things become interesting. You probably would want to use SVM where there a host of library you can use, e.g. libsvm, SVMlight and etc.
