Cauchy-Schwarz散度在图像分割中的应用

Application of Cauchy-Schwarz divergence in image segmentation

基于Cauchy-Schwarz散度,提出了一种新的主动轮廓线图像分割模型。该模型能量泛函有两部分组成:几何正则项和数据拟合项。其中,数据拟合项通过图像灰度的概率密度函数之间的Cauchy-Schwarz散度来加以构造,并且对概率密度函数进行了非参数估计。为了快速获得新模型的全局最优解,采用了模型的凸化及Split-Bregman快速迭代技术。通过一些图像的分割实验,验证了该模型可取得令人满意的分割效果且具有较快的收敛速度。

Based on the Cauchy-Schwarz divergence, a new image segmentation model using active contour is proposed, which consists of a geometric regularization term and a data fitting term. Particularly, the data fitting item is constructed by measuring the Cauchy-Schwarz divergence between the probability density functions of intensity in different regions, and the probability density functions are estimated with nonparametric method. In order to obtain a global and optimal solution for the new model quickly, the latest convexification technology and the Split-Bregman rapid iteration method are used. The experimental segmen- tation results for some images are demonstrated to show some desirable performances and faster convergence rate of this model.