Taking the derivative of an image is a concept that I’ve seen come up both in edge detection and in computing optical flow. It’s confused the heck out of me because I would normally think of derivatives in terms of taking the derivative of a continuous function. However, with an image, you have a 2D matrix of seemingly random values, so what could it mean to take the derivative?
Lecture 01 Introduction to Computer Vision
There is a copmuter vision lecture series with Dr. Mubarak Shah that the University of Central Florida recently published to YouTube. It looks like the lectures are from the fall / winter of 2012. I’ve found these lectures extremely helpful in a lot of the computer vision learning and research that I’ve been doing.
This tutorial covers SIFT feature extraction, and matching SIFT features between two images using OpenCV’s ‘matcher_simple’ example. It does not go as far, though, as setting up an object recognition demo, where you can identify a trained object in any image.
Finding simple setup instructions for getting some OpenCV sample code up and running is a pain. They seem to make significant changes in each release, which means that an article providing setup instructions for an older version may not work for the latest. A lot of the instructions are geared towards being setup to recompile OpenCV, which you’re probably not interested in when you’re just getting started.