多对象分割是图像处理中的一个难题,基于Multiway cut的图像分割是一种人工交互式多对象分割方法,能够实现图像的粗分割和精确分割。使用分水岭分割图像,把图像分割为属性相似的小区域;根据交互建立节点层次图,构建带权无向网络;不同层次的节点参与不同的运算,采用Multiway cut迭代分割;交互和分割可以多次执行,直至满足用户的要求。实验结果表明,该方法人工参与方便,准确度得到提高,速度满足现场操作的要求。
Multi-object segmentation is a difficult problem in image processing. Image segmentation,based on Multiway cut, is an interactive multi-object image segmentation,which can implement segmentation from coarseness to accurateness. Firstly, used watershed to segment a image into small regions with the same attributes. Then built a node hierarchy and a weighted undirected network based on interactions. Used the nodes of different hierarchy play different roles and the algorithm of Multiway cut to segment image iteratively. Interaction and segmentation could be executed many times until user’s requirements were met. Experimental results show that the method can reduce manual participation,effectively improve accuracy and its speed can meet the requirements of field operations.