Geodesic geometric flow for medical image segmentation based on similarities of interfacial gradients

For 3D medical images with strong similarities of interfacial gradients, a method is presented to bring local regional characteristics of segmented sections into adjacent sections waiting for segmentation and to guide the contour curves of the later to converge to actual boundary. The proposed method improves the stopping criterion of curve evolution through introduction of adjacent layer's prior information into edge detection function, so that introduces similarities of interfacial gradients into geodesic geometric flow as prior information to modify the leakage phenomena that exists in geometric active contour model when disconnected or weak edges appear in layers and improves the precision and stability of segmentation for 3D medical images.