中文摘要：

随机游走算法（Random Walk）是近来提出的一种基于图论的交互式分割算法,针对Random Walk算法计算耗时长的缺点,提出一种新的改进型算法。首先,该方法增加了分水岭算法（Watershed）作为预分割,有效地减少了后续分割算法的计算量。其次,根据分水岭预分割后的区域,重新构造连接域图,并采用K均值算法将预分割后的区域像素值进行聚类,同时利用聚类后的像素距离差值的高斯函数2范数作为能量函数。最后,用Random Walk算法对其进行求解。该方法比原算法速度更快,精度令人满意,并且可以进行多目标分割,对于噪声有很好的鲁棒性。

英文摘要：

Random Walk is an interactive segmentation algorithm which has been proposed recently. To solve the problem that it takes long time to calculate, this paper proposed an improved Random Walk algorithm, which not only provides an instant visual feedback, but also has the robust property. First, in order to reduce the computation cost, the Watershed transform was used for partitioning the image into small region pieces. Secondly, the image region graph was constructed based on the Water- shed pre-segment region with the typical Gaussian weighting energy function. Finally, the Random Walk algorithm was applied to obtaining the final segmentation result. The experiments show that this proposed method can achieve high segmentation accuracy and run much faster.