基于多级协同传播的交互式图像分割

现有的交互式分割算法对用户标记的数量与位置较为敏感，为此，提出一种基于两级协同扩散的分割算法。将均值漂移分割的小块区域作为虚拟像素，参与构建两级加权图，为信息向远方同质区域传播提供路径。采用状态迭代过程实现信息传播，收敛后得到所有像素对各种标志的隶属度，并进行隶属度比较，完成分割。实验结果表明，与单级传播算法相比，该算法即使采用少量用户标记也可以获得满意的分割结果，对标记的数量与位置具有较高的鲁棒性。

Interactive Image Segmentation Based on Multi-level Cooperative Propagation

Existing interactive segmentation algorithms are sensitive to quantity as well as placement of user’s scribbles, hence a novel algorithm which employs region cue is proposed. Segmentation result by mean-shift algorithm is regarded as imaginary pixels, and further incorporated into construction of two-layer weighted graph, which provides long range connection between homogenous semantic areas. Then all these connections are integrated in a chrominance iteration formulation, which performs updating to membership of all pixels to every label. Segmentation is completed via comparison to resulting membership. Experimental results indicate that even little amount of scribble can result in satisfactory result, which demonstrates that the proposed method has higher robustness to quantity and placement of scribbles in comparison with existing mono-layer ones.