Automatic object recognition applied to Where's Waldo

Abstract

The illustrated series of Wheres Waldo books [1] provide an excellent set of challenge problems for designing and testing hardware/algorithms to approximate human object recognition capability. Developing a process for automatically locating and identifying Waldo in his many varied environments incorporates aspects of pattern recognition, machine learning, and image processing. This paper addresses the development and evaluation of multiple algorithms used separately and in conjunction to find Waldo.