基于视觉注意的随机游走图像分割

发布时间：2018-06-11 20:03

本文选题：图像分割 + 随机游走　；参考：《仪器仪表学报》2017年07期

【摘要】：传统随机游走图像分割需要多次交互设置种子点以获得理想的分割结果。在视觉注意的基础上,提出了一种新的自动确定种子点的随机游走图像分割算法。首先对图像进行超像素分割,并生成概率边界图(PBM);然后基于Itti模型,通过视觉注意焦点的转移搜寻待分割的关键区域;为确定关键分割区域种子点,以当前注意焦点作为极点对概率边界图进行极坐标变换,在获得的极坐标概率边界图上建立关于焦点区域边界的能量函数,采用图论max-flow min-cut算法最小化能量函数检测焦点区域的最优边界,焦点区域边界内的超像素即为种子点;最后以超像素为节点构造图,在图上随机游走完成图像分割。在Berkeley Segmentation Data Set上的实验表明本文方法能有效分割复杂图像。
[Abstract]:Traditional random walk image segmentation requires multiple interactive seed points to obtain ideal segmentation results. On the basis of visual attention, a new random walk image segmentation algorithm is proposed, which can automatically determine seed points. Firstly, the image is segmented by super-pixel, and the probabilistic boundary map is generated. Then, based on the Itti model, the key areas to be segmented are searched through the shift of visual focus. Taking the current focus as the pole, the polar coordinate transformation of the probabilistic boundary map is carried out, and the energy function about the boundary of the focal area is established on the obtained polar coordinate probabilistic boundary map. The algorithm of graph theory max-flow min-cut is used to minimize the energy function to detect the optimal boundary of the focus region. The super-pixel in the focus region is the seed point. Finally, the super-pixel is used as the node to construct the graph, and the image segmentation is completed by random walk on the graph. Experiments on Berkeley Segmentation data set show that the proposed method can effectively segment complex images.
【作者单位】：西华大学机械工程学院;
【基金】：教育部春晖计划项目(12202528) 西华大学重点项目(Z1120223)资助
【分类号】：TP391.41

【相似文献】