边界与区域相结合的目标提取算法

Object extraction algorithm combining boundary information and region information

对于目标中带孔或目标颜色信息和背景颜色信息相似的图像,单独根据边界信息或区域信息无法提取出准确的目标。从图像的边界信息与区域信息综合考虑,以图割理论为基础,提出了一种基于图割的边界信息与区域信息相结合的目标提取算法。实验结果表明,该算法对目标中带孔或目标颜色信息和背景颜色信息相似的图像,能够准确、有效地提取出目标,自动分割精度高,用户工作量少,分割效率高,达到了预期的效果。

For some objects with holes, or their color information similar to the background＇s, object can＇t be extracted exactly only by boundary information or region information. Combining boundary information and region information simultaneously, an object extraction algorithm based on graph cut was proposed. The experimental results show that the algorithm can extract such objects exactly and effectively. The method achieves the expectation with high accuracy in automatic segmentation, less user＇s workload and high efficiency.