Abstract

One of the greatest abilities of the human eye is its capacity to perceive depth, an essential skill that allows us to perform fundamental tasks, such as avoiding obstacles and retrieving objects, as well as complicated tasks, such as driving a car. As advancements in the field of robotics allow robots to successfully perform these afore-mentioned tasks, the need for simulated depth perception, ideally in an efficient manner, continues to grow. With the specific application of creating an efficient depth finding algorithm for robots with simple binocular cameras, various optimizations, introduced by Pedro Felzenszwalb of the University of Chicago, were applied to a naïve belief propagation algorithm to achieve more efficient belief propagation in depth finding. This paper provides an overview of the naïve belief propagation algorithm, the algorithm optimizations, and experimental results and analysis on the impact of these optimizations on algorithm performance.