参数化形态学梯度修正的水平集肝肿瘤分割

Level set liver tumor segmentation based on parameterized morphological gradient modification

针对单一水平集算法处理低对比度或边缘模糊肝脏CT图像时,在梯度局部极小值区域或虚假边缘处常常会出现曲线停止演化现象的问题,提出了一种参数化形态学梯度修正的水平集图像分割方法进行研究。首先对图像进行形态学梯度变换,增强图像的对比度;然后以此为基础,在特定邻域内建立结构元素半径与梯度级的函数关系对图像进行梯度修正,增强目标边缘聚合度并去除图像噪声及非规则细节引起的局部极小值,同时减小目标轮廓位置的偏移;最后根据图像梯度信息运用水平集方法实现图像中单个或多个目标分割。实验结果表明,该算法有效地解决了标准水平集分割方法中存在的伪分割问题,能够对肝脏肿瘤进行较准确分割。

In dealing with low contrast or borderline blurred liver CT images with traditional level set algorithm, the evolving curve often stopped in local gradient minimal regions or false edges. In order to solve this problem, this paper proposed a level set segmentation method based on parameterized morphological gradient modification to study. Firstly, it transformed image＇ s morphological gradient to enhance its contrast. Then further it set up function relationship between structural elements radius and gradient within specific neighborhood and modified image gradient, therefore, to enhance the polymerization degree of the target edge, removed the noise and local minimal caused by irregular detail and reduced the migration of object＇ s contour position at the same time. Finally,it used level set method to segment single or multiple targets on consideration of the image gradient information. The experimental results show that the algorithm effectively solves the problem of pseudo edge segmentation in standard level set method. The liver tumor can be segmented with this algorithm accurately.