Active contours with prior corner detection to extract discontinuous boundaries of anatomical structures in X-ray images

Abstract

Attached

Active contours are a form of curves that deforms according to an energy minimising function and are widely used in computer vision and image processing applications to extract features of interests from raw images acquired using an image capturing device. One of the major limitations in active contours is its inability to converge accurately when the object of interest exhibits sharp comers. In this study, a new technique of active contour model to extract boundaries of objects having shaip comers is presented. By incorporating a priori knowledge of significant comers of the object into the deforming contour, the proposed active contour is able to deform towards the boundaries of the object without surpassing the comers. The ability of the new technique to accurately extract features of interest of anatomical structures in medical X-ray images having sharp comers is demonstrated.