中文摘要：

均值偏移算法是一种统计迭代算法，因为其具备良好的鲁棒性，所以被广泛地应用于计算机视觉与模式识别等领域。然而该算法因计算量大、收敛速度慢而无法适用于一些对实时性要求较高、资源受限的场合。提出一种改进的迭代算法，该迭代算法通过使用偏移均值邻近的样本点来代替它，进而在样本集中构建出迭代路径。相对于传统的均值偏移算法，该改进方法在不影响结果的情况下减小了算法复杂度。通过大量的聚类分析和图像分割实验对算法的有效性和普适性进行了验证。

英文摘要：

Mean shift is a statistical iterative algorithm. Because of its robustness, it is widely used in computer vision and pattern recogni- tion. However, the mean shift procedure has relatively high time complexity and slow convergence speed, so it can not be used in some spe- cial situations requiring high real-time property and unlimited resource. In this paper we present an improved iterative algorithm to replace it by using the nearby sample point of the shift mean value, and then constructs an iterative path in sample set. The improved algorithm reduces the time complexity under the condition of the result unaffected relative to traditional mean shift algorithm. Through a large number of experi- ments in cluster analysis and image segmentation we verify the effectiveness and universality of the proposed algorithm.