针对已有的基于分形维数的图像分割算法难以快速计算一个小区域的分形维数，计算复杂，效率低的问题，通过分析云的分形特征，提出一种采用局部分形维数的方法对红外云彩图像进行分割。首先，提出了一种高效算法来计算一段区间内的分形维数，使用树状数组作为数据结构，利用已经计算出来的信息，在O(logN)的时间内得到结果；然后，通过计算云图每个水平线的分形维数，将分形维数超过一定阈值的区域确定为云彩区域；最后，将每条水平线的高维数区间结合在一起得出整个分割结果。实验结果表明，该方法解决了传统的分形维数算法在大量计算维数时算法复杂度高、计算时间长的问题，对于640 pixel×480 pixel的大型图像计算时间<0.1 s；同时该算法能够有效地将云彩和其他人工遮挡物以及背景光线变化和局部噪声区分开，取得满意的处理效果。

An efficient image segmentation method based on local fractal dimension was proposed by analyzing the fractal characteristics of clouds to solve the problems existed in image segmentation algorithms in high complexity,low efficiency and difficult to calculate the fractal dimension of a small area quickly. Firstly, an efficient algorithm to calculate the fractal dimension of a small area was proposed. By utilizing a tree array as data structure and taking the advantage of calculated information, the result could be obtained in the time of O(logN). Then, through calculating the fractal dimension of each horizontal line in the infrared cloud image, the fractal dimension that exceeds a certain threshold area was identified as the cloud area. Finally, the segmentation result was gotten through combining all the high dimensions of each horizontal line. The result demonstrates that this method solves the problems of complexity and inefficiency of the traditional methods based on fractal dimension, and the computing time remains less than 0.1 s for a large image of 640 pixel×480 pixel. The algorithm can effectively separate the clouds from the other artificial objects, changed background light, and the local noise to obtain good segmentation results.