Overall, we are satisfied with the results of the project. We achieved our main goal of using a matched filter with color images, and we extended the techniques of 2D image recognition develeped by the DSP Pictomaniacs. Our technique of matched filtering allowed us to successfully match color images rather than simply the greyscale images that the DSP Pictomaniacs analyzed. From these images, we saw the benefits of image matching with color. It seems logical that adding the dimension of color to image matching would improve results, but we cannot say for sure because we never implemented our algorithm with greyscale versions of the same Waldo images. If we had done this, we could have made a better comparison of greyscale and color. We might have seen, for example, that adding color prevents matching the Waldo head with other similar images such as an Odlaw head or a Wenda head.
As mentioned in the Errors and Problems section, we also saw many drawbacks to our method. Using a generic Waldo head, for example, didn't always return the correct location. This was due in part to the variation between the different Waldo heads in each picture. We could have avoided this problem by using a library of Waldo heads that could be used to not only pinpoint Waldo's location, but also to find the best fitting Waldo head. If we had had more time, or if we were to extend this project, we would definately explore these different aspects of finding Waldo.
By the way, all Where's Waldo images and names are property of Martin Handford. And the real name of this project is No Waldo Here, Just My Balls