一种融合区域生长与图论的图像分割方法

An Image Segmentation Approach by Combining Region Growing and Graph Theory

该文提出一种融合区域生长与图论的图像分割方法,一般的基于区域的分割方法在区域生长完成之后需要进行区域的合并,以消除过分割现象。该文的方法在区域生长完成之后,用NormalizedCut方法在区域之间进行分割,产生最终所分割的图像。在方法上区域生长方法考虑的是图像的局部信息,NormalizedCut方法考虑的是图像的全局信息,该文的方法融合了两者的优点。该文的算法主要以灰度图像为研究对象,实验结果表明可以取得很好的分割效果。

This paper suggests an approach for image segmentation by combining region growing and graph theory.In general,approaches of region-based segmentation always carry out region merging to avoid over-segmentation after region growing.The approach of this paper uses Normalized Cut to segment between regions after region growing,and then produces the final segmented images.Region growing method focuses on local variations of an image,while Normalized Cut method can extract a global property of an image.Authors' approach combines both advantages.The algorithms of this paper mainly concentrate on gray level images.The experiment shows good results of segmentation.