【摘要】 图割算法是图像分割中经典有效的算法,针对其在前景/背景颜色有重叠时容易产生分割错误、shrinkingbias现象及交互实时性不佳的问题,提出一种利用视觉显著性与图割的交互式图像分割算法.首先利用Mean Shift算法将原始图像高效地预分割为基于区域的图结构,使得计算量大大下降;然后结合图像内容的显著性分析提高数据项约束的可靠性,并结合局部自适应的正则化参数,有效地改善了shrinking bias现象.实验结果表明,该算法交互快速,分割结果更加精确.更多还原

【Abstract】 Graph Cuts algorithm is a classical and effective method for image segmentation.However,it tends to produce segmentation errors and shrinking bias when the foreground and background color distributions overlap,and its interactivity is less efficient.To address these problems an interactive method for image segmentation based on visual saliency and graph cuts is proposed in this paper.First the Mean Shift algorithm is applied to efficiently pre-segment the original image into regions to construct region adjacency graph,which can greatly reduce the computational complexity.Then with image saliency analysis the reliability of the data item constraint is enhanced,and by using local adaptive regularization parameter the shrinking bias problem is improved effectively.The experiments demonstrate the superior performance of the proposed method in terms of interactive efficiency and segmentation accuracy.更多还原