An Improved Neuron Segmentation Model for Crack Detection – Image Segmentation Model

It is still very challenging to establish a unified and robust framework to perform accurate and complete crack extraction from images with cluttered background, various morphological differences and even with shadow influence. In this paper, an improved neuron segmentation model with two stages is proposed for crack segmentation. Firstly, a robust crack indicator function is designed based on local directional filtering; it makes up for the traditional function based on hessian matrix, which is resulting in problem of local structure discontinuities. After obtaining the indicator function, the crack detection is performed in an integrated mode; it is incorporating the automated directional region growing without manual intervention by adopting level sets; then efficient and complete crack segmentation is realized by iterative contour evolution. The performance of the proposed model is demonstrated by experiments on three kinds of grouped crack sample images and the quantitative evaluation. We also argue that the proposed model is applicable for biomedical image segmentation.