基于超像素的Graph-Based图像分割算法

Graph-Based Image Segmentation Based on Superpixels

针对EGBIS分割算法中的过分割问题,提出了一种基于超像素的graph-based图像分割算法SGBIS.首先,对图像进行基于简单线性迭代聚类（SLIC）的超像素预分割;然后以每个超像素作为节点构造带权无向图,以相邻超像素颜色平均值的欧式距离作为图中边的权值;最后利用基于图的算法合并超像素得到分割结果.用VI、PRI和F值3个指标分析了算法性能,结果表明,新算法可以得到更为理想的分割效果;引入交互分割区域合并,也可满足用户图像分割的需求.

As graph-based algorithm is inclined to over-segment,a new graph-based image segmentation algorithm based on superpixels called superpixel graph based image segmentation（SGBIS） was proposed and simple linear iterative clustering（SLIC） superpixels segmentation was employed as pre-segmentation. Then,the weighted undirected graph regarding superpixels as nodes was constructed,the Euclidean distance of adjacent superpixels＇ average color is used as the weight. Finally,the segmentation results are obtained by merging superpixels based on graph-based algorithm. Three indexes variation of information（VI）,probabilistic rand index（PRI） and F-measure are introduced to evaluate algorithm. Experiments show that it can get better segmentations. An interactive region merging interface is also introduced,which could meet users need very well.