Fuzzy c-means clustering based active contour model driven by edge scaled region information. (15th April 2019)

• Record Type: Journal Article
• Authors: Soomro, Shafiullah; Munir, Asad; Choi, Kwang Nam
• Is Part Of: Expert systems with applications. Volume 120(2019)
• Page Start: 387
• Page End: 396
• Publication Date: 2019-04-15
• Subjects: Image segmentation -- Active contour -- Level set; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2018.10.052
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Edge scaled method with region based statistical information is proposed. Global and local terms fused with compacted geodesic edge term in additive fashion. Proposed method is automated by FCM (fuzzy-c-means) clustering. The experimental analysis is performed on synthetic and real inhomogeneous images. Abstract: This research article proposes a novel edge scaled method with local and global region based statistical information for inhomogeneous image segmentation. We coordinate region force (local and global) term with geodesic edge term in level set formulation. The vital energy of the proposed model utilizes both global and local terms fused with compacted geodesic edge term in an additive fashion, which uses image gradient information to segment feeble boundaries inside images. The initialization of the level set is always vulnerable and its execution is liable to suitable initialization, which requires manual intercession. In this regard, the proposed method is extended and automated by integrating the proposed method with FCM (fuzzy c-means) clustering. Modified active contour method gets suitable initialization from FCM clustering, which eliminates the initial contour problem existed in the traditional region-based active contour methods. A development in this paper is to interface FCM with level set strategy by morphological tasks. Moreover, the uninterrupted segmentation scheme confirms the effectiveness of the proposed method in modern intelligent and automatic systems. The experimental analysis is performed on synthetic and real inhomogeneous images. The visual and quantitative results confirm the proficiency of the proposed technique and show that proposed method achieves better accuracy results compared to previous methods.