结合全局与局部信息的主动轮廓分割模型

摘要

Aimed at the problem of the traditional region-based active contour model, which was difficult to segment the noisy images and intensity inhomogeneous images, an active contour model, including global term, local term and regularization term, was proposed.Global term was derived from the data fidelity items of Chan-Vese model.Meanwhile, the construction of local term took the local intensity information into account and the image local entropy that reflected the grey characteristics was introduced.According to the characteristics of different images, the reasonable parameters of global item and local item were selected.The regularization term was added to ensure the smoothness of curve evolution and to guarantee the reliability of segmentation.A variable level set was used to minimize the energy function to get the gradient descent flow. Experimental results of synthetic images and real images demonstrate that the proposed approach is efficient in segmenting the noisy images and inhomogeneity images.%针对传统的基于区域的主动轮廓模型在分割灰度不均匀图像和噪声图像存在效果不佳的问题,提出结合全局项与局部项的主动轮廓分割模型.全局项由CV(Chan-Vese)模型的保真项构成,局部项的构建考虑局部区域信息的同时引入反映图像灰度特性的局部熵信息.依据图像灰度的特点,选择合理的全局项和局部项参数,并加入正则项保证曲线在演化过程中保持平滑,保障分割结果的可靠性.通过变分水平集方法最小化能量泛函,依据梯度下降流迭代更新水平集,完成曲线演化.采用模拟图像和实际图像进行实验分析,结果表明,所提出的结合全局项和局部项的主动轮廓模型可以高效地分割噪声严重以及灰度分布不均匀的图像.