一种SVM与区域生长相结合的图像分割方法

Interactive segmentation method with SVM and region growing

作为一种全局门限处理方法,支持向量机图像分割方法不能完成对图像进行精细分割,其分割结果需要其他分割方法进一步处理。提出一种结合支持向量机和区域生长的交互式分割方法,不仅可有效剔除与感兴趣区域特征类似的非目标区域,而且把为SVM选择训练样本和为区域生长选择种子点两个步骤合二为一,从而提高了图像分割质量和交互式分割方法的自动分割能力。

The SVM segmentation method, as a kind of global thresholding, cannot segment the image finely. The results need to be post-processed by other segmentation approaches. In this paper, a new interactive segmentation method was proposed, which integrated the SVM and region growing. The proposed method can remove the regions of non-interest that have similar characteristics to regions of interest. Furthermore, it can combine the two procedures into one： the procedure of selecting samples for training SVM and the procedure of selecting seeds for region growing. As a result, the method improves the quality of segmentation and the performance of auto-segmentation.