A Robust Image Segmentation Model Based on Integrated Square Estimation

Shuisheng Xie, Jundong Liu, Darlene E. Berryman, Edward O. List, Charles D. Smith, Hima Chebrolu

Abstract: This paper presents a robust segmentation method based on the integrated squared error or L 2 estimation (L 2 E). Formulated under the Finite Gaussian Mixture (FGM) framework, the new model (FGML2E) has a strong discriminative ability in capturing the major parts of intensity distribution without being affected by outlier structures or heavy noise. Comparisons are made with two popular solutions, the EM and FCM algorithms, and the experimental results clearly show the improvement made by our model.