Two dimensional active contour model on multigrids for edge detection of images

Abstract

Low-level tasks have been widely regarded as autonomous bottom-up processes in computational vision research. Examples of low-level tasks are edge detection, stereo matching, and motion tracking. In medical imaging, active contours have also been widely applied for various applications. In fact, active contours is one of the most popular PDE-based tools and powerful tool in performing object tracking. Active contour model, also called classical explicit snake was first introduced by Kass, Witkin and Terzopoulos. The main weaknesses of this method relate to not only the intrinsic characteristics of the contour, but also the parameterization, in which it is unable to handle topological changes. To solve these problems, a different model for active contours based on geometric partial different equation is proposed which is independent of parameterization, intrinsic and very stable. The important development has been the introduction of geodesic active contours. Levelset method was introduced for the moving fronts capture, where the active contour method is given implicitly as the zero level-set of a scalar function defined on implementing the entire image domain. This allows for a much more natural changes in the topology of the curve than parametric snakes. However, the main weakness of level set methods is that the complexity of the computational cost is high. A fast algorithm using semi-implicit addictive operator splitting (AOS) technique is used to restrict the computational cost. Edge detection based on semi-implicit is implemented for the edge detection on medical images such as medical resonance image (MRI). Multigrid is a numerical method that has a good accuracy and stability even with big time step. Exploiting these properties, multigrid was adopted for implementation of the geodesic active contour model. MATLAB has been chosen as the development platformfor the implementations and the experiments since it is well suited for the kind of computations that are required. Besides it is widely used by the image processing community. Experimental results demonstrate the multigrid is the most appropriate method that can applied with AOS implementation for medical imaging to detect the location of the tumor which can decrease number of iterations.