基于双区域演化水平集的图像分割方法

Double regional evolution based on level set for image segmentation

针对图像由于光照不均引起的灰度不均匀导致图像精确分割困难的问题,提出了一种双区域水平集演化(DRE)方法用于图像分割。双区域水平集演化方法介绍了局部区域控制项和矩形的初始化轮廓,能够快速地分割灰度不均匀图像,加快曲线演化速率。为进一步消除噪声对图像的影响,分割方法采用高斯滤波算子和卷积计算,特别是提出的方法使用一种新的隐含的惩罚能量函数作为正则化项,该项嵌入到水平集演化公式里,增加了曲线演化梯度流的约束的条件,影响水平集演化曲线向双边扩展,这样提高了图像轮廓划分的计算效率和精度。通过实验结果对比,提出的水平集演化算法能够较好地划分图像轮廓,得到精确的图像分割结果。

This paper concerned with the issue of image segmentation for intensity inhomogeneity ,and a novel double regional evolution （DRE） based on level set method is proposed for image segmentation .The proposed DRE method introduces the local regional control term and the rectangle initialization contour ,which can able to quickly segments uneven gray‐scale images and accelerates the curve evolution rate .In order to reduce the noises effects for image ,the image segmentation method adopts Gaussian filter operator and convolution calculation .In particularly ,DRE method develops a new potential penalty function as regularization term , which is embedded into the level set evolution equations to increases the constrained conditions based on the gradient flow conditions .Due to the potential function promotes the image contour evolutions with bilaterally extended , during the evolution , it is improved the image contours segment efficiency and numerical accuracy .Compared with the recent proposed methods based on level set , there are some superiorities of DRC method for image segmentation .