中文摘要：

图像分割作为从图像中提取感兴趣对象的必要步骤，通常需要其能够提供多尺度的分割结果．提出一种基于二叉划分树（BPT）的多尺度图像分割算法，用于系统地记录从图像的任意过分割结果上进行的区域合并过程；然后基于BPT中每个结点与其父亲结点的差异度量，提出一种包括自动确定候选结点和分裂合并策略的结点选择算法，来选出符合期望分割区域数目的结点，并生成相应尺度的分割结果．实验结果表明，文中算法能够在较粗分割尺度下获得更适合于对象提取的分割结果，有助于提高自动对象提取的效率以及减少交互式对象提取中的人工交互．

英文摘要：

As a necessary Stepfor interested object extraction from images, an image segmentation algorithm usually needs to provide a multiscale segmentation result. This paper proposes a binary partition tree （BPT） based multiscale image segmentation algorithm. Starting from an oversegmentation result, region merging is performed and the merging sequence is recorded by a BPT. Then each node in the BPT is measured with the difference from its parent node, and a node selecting algorithm combining automatic candidate node determination and splitting-merging scheme is proposed to select a subset of nodes, which are used to generate the segmentation result under any scale specified by the desired number of segmented regions. Experimental results demonstrate that the proposed algorithm can obtain better segmentation results suitable for object extraction under coarser scales, and thus improves the segmentation efficiency in automatic object extraction and reduces the user interaction activities in interactive object extraction.