基于全散度的C-V模型阈值法

Threshold method of C-V model based on total Bregman divergence

为了提高传统C-V模型的收敛速度并降低其对噪声的敏感性,提出基于全散度的C-V模型及其快速阈值分割算法。将Bregman散度与全散度统一获得新的全散度,并将其引入C-V模型的拟合偏差项,提高图像灰度值与分割区域平均灰度偏差值计算的鲁棒性。同时,采用变分水平集理论获得基于直方图的快速阈值计算方法。实验结果表明,该方法分割效果及收敛速度得到提高,且具有较好的鲁棒性和抗噪性。

The convergence rate of the traditional C-V model is slow,and it is sensitive to noise.The C-V model based on total Bregman divergence （TBD） is proposed and its corresponding fast threshold segmentation algorithm is implemented.Parametric TBD is built for unifying Bregman divergence and TBD,and utilized to rebuild the fitting iterms of C-V model.Meanwhile,the numerical algorithm relating to grayscale histogram is realized under the variational level set method.The experimental results show that the proposed algorithm illustrates excellent performance to natural images.It improves the convergent speed and has higher robustness and anti-noise features.