一种新的基于相位信息的活动轮廓模型
摘 要 传统的基于边缘检测的几何活动轮廓模型利用图像梯度信息进行边缘检测，对图像噪声极其敏感， 对弱边缘的捕捉能力较差，容易造成边缘泄露。本文提出的模型采用了一种新的基于图像相位信息的边缘检测函数，并且加入了一个额外的边缘定位能量项。与传统模型相比，本文模型对噪声的抑制能力得到加强，同时具有较强的弱边缘捕捉能力，无论对于普通图像还是超声图像，都能达到较为理想的分割效果。 我们利用本文提出的模型对普通图像和超声图像分别进行实验，并与当前几种主流的模型进行了对比。 结果表明，本文提出的模型具有传统模型不可比拟的一些优越性能。
关键词 活动轮廓模型； 水平集；相位一致性；边缘检测函数
A Novel Active Contour Model Based on Phase Information
Abstract Traditional geometric active contour models based on edge detection use gradient information to detect image edges, and are extremely sensitive to image noise. Their capability of capturing weak edge is poor, and easy to cause the leak of edges. The proposed model uses a new phase-information based edge detection function, and adds an extra edge alignment energy term. Compared with traditional models, its noise suppression ability has been strengthened , at the same time, it has strong capability of capturing weak edges in the image , both for ordinary images and ultrasound images, it can achieve ideal segmentation results. We use the proposed model to experiment on ordinary images and ultrasound images respectively, and we compare our model with several mainstream models. The results show that the proposed model has some unmatched advantages which traditional models don’t have.
Keywords Active contour model; Level set; Phase congruency; Edge detection function
