一种改进局部二元拟合的灰度非均匀图像分割

针对局部二元拟合(local binary fitting,LBF)能量模型对活动轮廓曲线初始位置较为敏感的缺点,本文提出一种改进局部二元拟合的灰度非均匀图像分割模型。首先把水平集函数初始化为一个常数,然后在迭代过程中引入一个扰动项,从而引导目标区域的水平集函数值发生符号变化。实验结果表明,上述模型能够有效地应用于灰度非均匀图像的分割。与LBF模型相比,本文模型无需人工选择活动轮廓曲线的初始位置,且避免了由于初始位置选择不当造成的分割错误。

An Improved Local Binary Fitting Model For Non-Uniform Image Segmentation

In order to overcome the shortage of Local Binary Fitting model＇ s sensitivity to initial contour position, this paper develops an improved local binary fitting （LBF）model for non-uniform image segmentation. Firstly the level set function is initialized to a constant and then the disturbance term is introduced into the iteration, making the sign of function value in object area change from negative to positive. Experiments show that our proposed method is able to work effectively on segmentation of non-uniform images. Compared with LBF method, our improved LBF method does not need to artificially select the initial position of active contour curves and avoids the segmentation error caused by improper choice of initial contour position.