Adaptive image segmentation based on local neighborhood information and Gaussian weighted Chi-square distance

We propose a novel affinity matrix for image segmentation in this paper. The affinity matrix is constructed by using the Gaussian weighted Chi-square distance with neighborhood information, in which the vital spatial structure of the image is considered. An adaptive local scaling parameter is used to refine the segmentation rather than selecting a single scaling parameter. We demonstrate that graph-based segmentation approach to solve the automatic segmentation of the vertebral bodies from sagittal magnetic resonance images of the spine is directly impacted by the affinity matrix. The encouraging results indicate that our new algorithm has the advantage of high accuracy and strong robustness.