Research On PDE-based Active Contour

Medical image segmentation is the foundation of medical image analysis and understanding. It is a common that there are low-contrast, the weak edges and uneven distribution of gray-scale phenomena in medical images, the expertise in the medical field is often required to make a reasonable segmentation. Active contour model provides a mechanism for the introduction of high-level knowledge, and is one of the most successful theories in the medical image segmentation field.The basic idea of active contour model is as follows:first model the image, namely, construct a energy functional, and then minimize the energy functional through optimization techniques. Thus image segmentation problem comes down to solving nonlinear partial differential equations (PDE) problem. The initial contour is equivalent to initial solution of the PDE, the process of minimizing the energy functional is the process of deformation of the curve, the image segmentation process is finished when the energy functional to achieve the minimum. This paper studied the local fitting based LBF model; add global control information and the edge gradient control information to form a new GLE model. then propose a new method for the measure of the difference, which can accurately measure the shape of small differences, The combination of GLE model and the new shape energy function is a new prior segmentation model GLES.In addition, this paper also proposes a new algorithms for the generation of signed distance function and a pose initialization algorithm based on image moments.This effectively reduces the time required for the evolution of affine transform, accelerate the evolution speed of GLES model.The experimental results on a series of synthetic images and real medical images show that the proposed algorithm can obtain more accurate segmentation results in the case of uneven distribution of pixels and can effectively improve segmentation accuracy in real medical images.