基于曲面拟合的图像分割算法

传统分割方法,在光照不均匀情况下,很难得到理想的分割结果。针对这种情况,提出一种基于曲面拟合的阈值曲面分割方法。首先利用偏离项和光顺项构造拟合曲面方程,然后使用在统一的大光顺项因子条件下求解的拟合结果,来构造自适应的偏离因子与光顺因子,最后利用这些自适应因子第二次精确拟合阈值曲面。实验表明,该方法对于照度不均匀的图像,分割结果明显优于传统全局阈值法。

Image Segmentation Based on Surface Fitting

The traditional segmentation method cannot get the ideal result under the non-uniform illumination.With regard to this situation,a segmentation method using threshold surface based on surface fitting is proposed.Firstly,using deflection item and fairing item to construct the equation of fitting surface.Secondly,In order to construct the adaptive weight factors,using the result of fitting surface with more heavy fairing item,then fit the threshold surface exactly again.The experimental result shows that the method has better than traditional method.