摘要
针对Grab Cut基于像素建立图模型并进行迭代求解耗时的特点,提出了一种新的基于SLICO改进的Grab Cut分割新算法。首先用户在图像目标区域手动划定一个矩形框,然后在CIELab颜色模型下利用SLICO算法将图像预处理成内部颜色一致的超像素图,利用这些超像素来构建图模型,并用这些超像素均值迭代估计高斯混合模型(GMM)参数。在参数估计中,采用背景区域优化技术,显著减少迭代时的节点数量,并减少矩形框外颜色的干扰,最后利用最小割(min-cut)算法求得图模型的最优分割。实验结果表明了该算法在精度和速度上都有很好的性能。
To overcome the disadvantage of time load of Grab Cut＇s for the image segmentation that set up the graph model in pixels and processed it iteratively,this paper proposed a fast method based on image segmentation of Grab Cut that was combined with SLICO. Firstly,it calibrated a rectangular box in the target zone manually,then splitted the image into small areas of the similar colors named super pixels with SLICO in CIELab color model. This method used super pixels to set up the graph model and estimated GMMs iteratively. In addition,it used background optimization method to sharply decrease the node count of graph and the influence of colour out of target zone. Finally it used min-cut algorithm to get the optimal segmentation of graph. The experimental results show that the proposed method can significantly improve the performance in terms of accuracy and efficiency.
出处
《计算机应用研究》
CSCD
北大核心
2015年第10期3191-3195,共5页
Application Research of Computers
基金
国家交通部科技项目(2011318740240)
