一种全局隶属度和小波能量相结合的活动轮廓模型

计算机化X射线体层照相(computeried tomography, CT),图像中的磨玻璃型肺结节,由于其具有模糊的轮廓,且肺结节区域与其邻域的亮度值相差很小等特性,一般的图像分割算法很难对其进行精准地分割。针对该问题,提出一种全局隶属度和小波能量相结合的活动轮廓模型。首先,利用全局隶属度函数调整初始活动轮廓曲线,使之与目标边界的距离更接近,且形状更相似;同时,基于全局隶属度的边界停止函数能快速收敛于目标边界,使得预测曲线更加贴合待分割目标;其次,基于小波能量的局部活动轮廓模型数据项增强了目标与背景之间的对比度,进而能更精准地分割在低对比度和亮度非均匀场景中的目标轮廓。将该模型应用于全实质及部分实质磨玻璃型肺结节的CT图像中,实验证明了本文算法的优越性。

An Active Contour Model based on Global Membershipand Wavelet Energy

For the local region-based active contour models hard to accurately segment object contour in image which contain fuzzy boundary, low contrast and intensity inhomogeneity regions, in order to cope with these problems, an active contour model based on global membership and wavelet energy is proposed. First, the wavelet energy is incorporated into the active contour model. The wavelet energy can enhance the dissimilarity between the object region with background region, and the local region information is considered, so the proposed active contour model can segment low contrast region with intensity inhomogeneity. Second, the speed function based on global membership is incorporated into the active contour model. At the boundary of object region, the speed function of evolution equals to zero, so the active contour model can segment object region with fuzzy boundary. Third, the initial contour is set near the boundary of the object region using global membership, so the evolution curve may obtain a global minimum. Experiments in the segmentation of ground glass opacity (GGO) pulmonary nodules show the superiority of our method.