While the OpenCV library is interesting to use, it doesn't offer a lot of transparency as you learn. If you're interested in actually learning about the field, I would recommend looking into low-level image processing libraries and implementing your own Computer Vision applications. Once you've coded your own basic CV applications, using the OpenCV library becomes a lot easier. I would suggest the following topics to advance quickly through the basics:
- sobel operators for edge detection
- trying your hand at color segmentation
- reconstructing 3d information from stereo images using disparity maps
Here's a site with some good test images (http://www.cs.cmu.edu/~cil/v-images.html).
I also found a good resource of course slides that cover the majority of these topics at (http://www.cs.nott.ac.uk/~tpp/G5BVIS/lectures.html)
Happy hacking =)