复杂背景下基于level set模型的头部分割算法

在办公室环境下，往往出现头部因遮挡、光照变化难以分割的情况。针对此类问题，提出基于level set模型的头部分割鲁棒算法。用符号距离函数建立头部level set模型，将头部level set模型与待分割图像符号距离函数的差异项添加到Chan Vese分割方程中，建立新的Euler Lagrange方程。新的Euler Lagrange方程中符号距离函数进化终止，即为头部的分割结果。将提出的分割算法和Chan Vese分割算法用于两人遮挡、加高斯噪声、光照变化三种办公室背景情况下的分割，结果表明提出的分割算法对于以上三种背景情况下的头部分割是鲁棒的。

Robust head segmentation method based on level set model in complex scenes

In office setting,head segmentation becomes difficult in the scene with occlusion or frequent brightness change.To deal with such problems,a robust head segmentation method based on level set model was proposed.After head level set model was constructed by sign distance function,the difference between head level set model and sign distance function of an image was added to Chan-Vese segmentation formula.Then a new Euler-Lagrange formula was obtained.When the evolvement of the sign distance function in the new Euler-Lagrange formula stopped,the head segmentation was finished.The experiments were implemented separately in office setting with scenes of two-person occlusion,Gaussian noise,and brightness change.Compared with Chan-Vese segmentation,the provided head segmentation method is robust to above three complex scenes.