A medical image segmentation algorithm based on bi-directional region growing

Image segmentation is one of vital researching branches in medical image processing and analysis. Considering the characteristics of medical images, we propose a bi-directional region growing segmentation algorithm. The interests of the algorithm include the easiness of initial seed selection and robustness to noises and the order of pixel processing. This method also holds for other segmentation applications in which background region is simple but target region is complex. In order to select an appropriate threshold, the concept of Neighboring Difference Transform is proposed. The issue of threshold selection is converted to minimization problem with the assistance of statistical properties of transformation matrix. Experimental results show that the algorithm can accurately obtain medical image segmentation results.