基于先验形状信息的水平集图像分割

摘要

针对现有水平集方法对于具有强噪声或弱边界的目标进行分割时存在的问题,提出了一种基于形状先验的图像分割方法.该模型采用变分水平集方法,融合了区域特征和边界轮廓特征,并通过相似性匹配选择最佳先验形状.该模型不仅对具有强噪声和弱边界的复杂图像具有较好的分割效果,而且有效地解决了曲线演化的初始轮廓的确定问题.与传统方法进行对比实验,结果表明,该方法具有较好的分割效果和较高的准确率.%A shape-prior based on image segmentation method was proposed to approach the defect of the current level set method in segmenting the subject with strong noise and weak boundary. By using level set method, this model combines region and boundary information together and selects the optimal prior shape by Similarity Matching. The model displays its advantages in the segmentation of complicated image with strong noise and weak boundary,and in the efficiency to solve the problems of confirming the primary contour of curve evolution. Compared with traditional way, the experiment proves that this approach contributes better segmenting effect and improved accuracy.