融合双边滤波算子的多相水平集图像分割算法

由于乳腺肿瘤超声图像的边界模糊,且灰度异质现象较严重,准确分割出肿瘤区域是一项具有挑战性的工作.针对传统的Chan-Vese模型和局部二值拟合模型(local binary fitting)的分割缺陷,在乳腺肿瘤超声图像的全局和局部能量信息的基础上,结合双边滤波算子,提出一种全局和局部二值拟合模型的多相水平集分割算法.首先,将双边滤波算子作为乳腺肿瘤超声图像的核函数;然后,根据变分法求解表征超声图像结构信息的能量泛函,得到对应的梯度矢量方程;随后,引入多相水平集函数实现病灶区域的多区域细化分割;最后,对乳腺超声图像数据集的分割实验.结果 发现:经过与医生手动标记的肿瘤区域进行对比,分割准确度为94.51％.可见,该模型的准确度较高、误判率较低、鲁棒性较强.

Image Segmentation Algorithm Based on Multi-phase Level Set Combined with Bilateral Filter Operator

Because the boundary of breast tumor ultrasound image is fuzzy and the gray heterogeneity is serious, it is a challenging work to accurately segment the tumor region. Aiming at the defects of the traditional Chan Vese model and local binary fitting model, based on the global and local energy information of breast tumor ultrasound image, combined with bilateral filter operator, a multi-phase level set segmentation algorithm based on local and global binary fitting model is proposed. Firstly, the bilateral filter operator is used as the kernel function of breast tumor ultrasound image; then, the energy functional representing the structure information of ultrasound image is solved according to the variational method, and the corresponding gradient vector equation is obtained; then, the multi-phase level set function is introduced to realize the multi region thinning segmentation of the lesion region; finally, the segmentation experiment of breast ultrasound image data set is carried out. The results showed that the accuracy of segmentation was 94.51%. It can be seen that the accuracy of the model is high, the error rate is low, and the robustness is strong.