提出了一种基于分形特征分割自然景物图象的方法。分割的主要目标是找出景物中的人造物体。首先构造了一组分形纹理特征，给出了图象不同部分的粗糙度及纹理基地大小、方向特性的描述，以此为基础进行图象分割。此外，通过对盒维数方法中的参数拟合性能的分析，对维数估计方法进行优化，理论和实验表明，在自然环境中，该方法能够区 的纹理区域及人造目标，达到满意的分割结果。

In this paper, we proposed a fractal characters based method for nature environment image segmentation. The main goal of the segmentation is to find the artifact object from the image. Based on several kinds of fractal characters, we can acquire the description of roughness of each part of the image. Then we can get the segmentation of the image. We also analysis the performance of the data fitting in the box dimension estimation, and present an optimizedmethod. According to a lot of experiments, the proposed approach is proved to be suitable for texture segmentation and artifact object finding in natural environment images.