Improving Image Segmentation Algorithms with Differential Evolution

• Published in: Reliable and Autonomous Computational Science pp. 381 - 396
• Main Authors: Hung, Chih-Cheng, Ren, Ziwei, Li, Wenfan, Yang, Qing
• Format: Book Chapter
• Language: English
• Published: Basel Springer Basel 2010
• Series: Autonomic Systems
• Subjects: Cluster Center; Error Matrix; Image Segmentation; Particle Swarm Optimization; Satellite Image
• ISBN: 9783034800303; 3034800304
• DOI: 10.1007/978-3-0348-0031-0_20

This paper proposes three algorithms based on the K-means, the simple competitive learning (SCL) algorithm, and the fuzzy c-means algorithm with differential evolution algorithm for image classification. Due to the local optimal performance of these three algorithms, the differential evolution (DE) is integrated with the K-means algorithm, the SCL algorithm and the fuzzy c-means algorithm to avoid the local optimal solution and improve the performance of three algorithms. One is called the DE-K-means algorithm which is a combination of DE and K-means; one is called the DE-SCL algorithm which is a combination of DE and SCL; and the other DE-FCM algorithm which is a combination of DE and FCM. The preliminary experimental results show that these proposed algorithms are more promising on image segmentation.