Evaluation of active contour on medical inhomogeneous image segmentation

Abstract

Segmentation is an important step in medical image analysis. This process is crucial but challenging due to inhomogeneneity in intensity of images. In addition, the images are often corrupted by noise and with contrast edges. There are some approaches aiming to cope with this kind of images such as: region growing, region competition, watershed segmentation, global thresholding, and active contour methods. Among them, active contour methods, especially level set-based active contour is widely used for image segmentation by their advantageous properties such as topology adaptability, and robustness to initialization. In this paper, we present and demonstrate the effectiveness of some recently active contour models for segmenting medical images with inhomogeneity in intensity. Among these techniques, the local binary fitting based model is validated as a promising method for medical image segmentation.