基于几何形变模型的医学图像分割研究

Based on the Geometric Deformable Model of Medical Image Segmentation Research

基于形变模型的图像分割方法通常可以分为参数型(Parametric Deformable Models)和几何型(Geometric Deformable Models)两类。提出了一种基于几何形变理论的LBF模型。针对水平集level set模型不能处理灰度不均一图像的分割问题,采用了LBF模型,并且该模型引入了一个以高斯函数为核函数的局部二值拟合能量,以获取图像的局部信息。通过理论分析与计算机仿真算例和其他算法的性能进行对比,表明改进算法LBF提高了图像分割的稳定性和精确性,具有较高的实用价值和广泛的应用背景。

Based on the deformation model image segmentation method usually can be divided into parametric deformable models and geometric deformable models two kinds, a LBF algorithm based on region-based active contour model was presented. Aiming at overcoming the difficulties caused by intensity inhomogeneities, LBF introduce a Gaussian kernel function to define a local binary fitting energy, so that local intensity information could be embedded into a region-based active contour model. Through theory analysis and simulation examples about these algorithms, experimental results show that the LBF algorithm improves stability and accuracy of the image segmentation, which has high practical value and broad application prospect.