Poisson Gradient Vector Flow of Active Contour Model

• Main Author: 陳信亨
• Other Authors: 李林滄
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/66757343004854539143

碩士 === 國立中興大學 === 應用數學系 === 91 === Abstract The paper presents a new external force field for Active Contour Model (ACM). This external force field, which we call Poisson Gradient Vector Flow (PGVF), is a concept modified from Gradient Vector Flow (GVF). Image segmentation has two difficulties with the traditional ACM. The initial curve must be close to the boundary of interested objects or the result will be far away from our expection. This is the first one problem can not be conquer by ACM. The second problem is the concave boundary which the traditional ACM is impossible to capture the concave boundary of interested image region. Although the proposed PGVF method is different from the GVF method, they can both effectively conquer the two difficulties of the traditional active contour models. The main dfference between PGVF and GVF scheme is the method of constructing the image force field. The GVF method uses the absolute value of gardient image intensity as edge map and the PGVF method uses canny edge algorithm to compute the edge map. The GVF method must introduce the time variable to solve initial value problem and use it interactively to calculate the image force field from partial differntial equations. The PGVF method uses the Poisson equation and the boundary conditions to obtain the potential fuction. In the paper we use finite difference method to solve the boundary value problem instead of intial value problem. The PGVF method without considering time variable avoids the instability condition of initial value problem as discussed GVF. The image force field obtained by the negative vector field calculated from the gradient of the potential function is the same procedure in both the PGVF and GVF mothods. The PVGF external force field applying on synthetic and medical images to capture objects with variable boundary images is considered in experiments. The results show that PVGF can accurately capture object boundary as well as GVF and save more time than GVF.