A novel image segmentation algorithm based on multi-motive reinforcement learning and OTSU
Qiao Sun1, Feixiang Chen1 , Hui Han1, Fu Xu1, Yanan Shi2
COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(12B) 39-44
1 School of Information Science and Technology, Beijing Forestry University, Beijing 100083, China
2 College of Computer Science and Technology, Jilin University, Changchun 130012, China
Image segmentation is one of the key technologies of computer vision. Among the image segmentation algorithms, the threshold-based approach is a simple and effective one. OTSU algorithm is considered to be one of the best approaches for threshold selection, but its drawbacks are the high time complexity and poor real time capabilities. In order to solve this issue, an efficient image segmentation algorithm based on multi-motive reinforcement learning is proposed in this paper, in the framework of OTSU, multi-motive reinforcement learning algorithm is adopted to get the optimal threshold for image segmentation. The learning motivation and action for threshold learning are defined in this article, and the original State-Action dual-layer structure is extended to State-Motive-Action triple-layer structure. Compared to traditional approach, the proposed approach has more flexibility, and is easier to integrate priori knowledge. The experimental result validated the effectiveness of the proposed approach.