GrabCut是一种快捷准确的交互式图像分割方法.但是,当待处理图像复杂度较大时,用户很难有效的标注矩形框,而且运算时间较长.针对以上问题,提出了一种改进的GrabCut算法.该算法通过视觉显著性实现矩形框的自动标注,与超像素的结合有效的减少了分割算法的时间.首先,通过一种结合改进超像素的流形排序算法来得到显著性图,并进一步得到目标的矩形框,然后用改进的超像素来构建GrabCut图割模型,最后,进行参数迭代估计从而得到分割图像.实验表明,本文提出的方法在保证GrabCut算法精度的前提下,实现了自动分割,并有效的减少了分割时间.%GrabCut is a fast and accurate interactive image segmentation method. However, as digital image technology develops and becomes increasingly complex, not only is it difficult for users to effectively mark a rectangular box, but also the operation time is too long. To solve the above problems, this study implements automatic labeling of rectangular boxes by salience, and effectively reduces the time required to run GrabCut by combining it with superpixels. First, we obtain a saliency map using graph-based manifold ranking, and further get the target rectangular box. Then, the GrabCut graph model is constructed using the improved superpixels. Finally, an iterative algorithm estimates the parameters that are used to obtain the segmentation results. Experiments show that the proposed method can effectively reduce segmentation time and user interaction while maintaining the accuracy of the original GrabCut algorithm.
展开▼