Description

Abstract

Ø A novel FCM-based method for accurate and noise-robust image segmentation is proposed. The method integrates spatial contextual information of the neighboring pixel into the FCM algorithm using phase congruency, anisotropic neighborhood and adaptive weights for membership calculation.

Ø The proposed method is designed to accurately segment edges, lines and coarse homogeneous regions at the same time.

Ø  The quantitative and qualitative experimental results for synthetic and real images show a good segmentation performance, especially for very noisy images and demonstrate an encouraging future of practical applications of the proposed method.