Abstract
Reinforcement learning (RL) is an approach to machine intelligence. It combines the fields of dynamic programming and supervised learning to yield powerful machine-learning systems. The RL appeals to many researchers because of its generality. However, it has not been used yet in the field of image processing. In RL, the computer is simply given a goal to achieve. The computer then learns how to achieve that goal by trial-and-error interactions with its environment. Of the RL methods Q-learning is a typical learning approach. In this paper, we present a novel method for image segmentation based on the Q-learning. Additionally, we illustrate the proposed algorithm and demonstrate its effectiveness for image contour extraction and region-of-interest detection using three medical images. Our preliminary results are promising.
