Where's Waldo?

You of course have seen those wildly popular, yet often frustrating, picture books that are filled with hundreds of little people in a crowded scene. Your mission: find a little man in blue pants and a red-striped shirt with glasses and a stocking cap. (This is Waldo.)

Now imagine a computer program that will locate Waldo for you. Just scan in the picture, run it through the Image Pictomaniacs' code and Viola! You have found Waldo.

This would ruin the fun of the book, yet in other applications, pattern matching might be useful. For example, for visual detection in robotics, pattern matching may be used to respond to certain user-defined preconditions such as stopping at red lights, detecting signs, etc.

Implementation

We will be using digital signal processing techniques to locate patterns within digitized images. Using the FFT to implement a two-dimensional convolution, we can obtain a useful function which allows us to determine the existance and location of the desired template.
We plan to proceed in the following manner:

• develop a rudimentary understanding of Matlab image reading commands and 2-dimensional convolution methods.
• develop tests for basic pattern recognition such as a black box on a white background.
• create more sophisticated testing scenarios such as variable shapes and more complex images.