一种新的用于图像分割的活动轮廓模型

为了从灰度异质图像中更好地提取目标,本文提出了一种新的图像分割方法,采用测地时间函数作为局部二值拟合模型的核函数,并结合图像区域的局部灰度信息和全局灰度信息建立能量泛函.同时,符号函数惩罚项的引入避免了符号函数的重新初始化,而曲线长度调整项则保证了曲线演化的连续性和光滑性.通过变分水平集方法最小化新的能量泛函,得到曲线演化的梯度下降流.通过对医学CT图像进行分割实验,证明了该方法的可行性和优越性.

A New Active Contour Model for Image Segmentation

In order to extract the object boundaries from the images of intensity inhomogeneity,a new image segmentation method is proposed.The new energy functional is set up by local and global intensity information,and the geodesic time is defined as kernel function of local binary fitting model.For there introducing the penalty term of signed distance function,the reinitialization is no longer required.The arc length adjustment term has ensured the evolution with continuous and smooth curve.The gradient descent flow of curve evolution is obtained by energy minimization through the variation level set method.The feasibility and robustness of the proposed model had been certified by the experiments of medical CT images segmentation.