中文摘要：

分数维作为一种图像纹理特征常用于遥感图像分类，提高图像分数维的测量有助于提高遥感分类精度。本文在介绍目前常用的基于分形布朗运动分数维计算方法的基础上，运用分形的相关理论，指出了现有方法存在的缺陷，提出了改进的分数维计算方法；并用SPOT全色影像中常见的5种地物进行了对比实验，实验结果表明，本文提出的改进方法获得的分数维测量结果比现有的方法更接近理论分析结果，更有利于地物的识别与分类。

英文摘要：

Fractal dimension, as a kind of texture feature, is often used for classification. It is important for classification to improve the precision of the fractal dimension. The current method of fractal dimension of image was introduced in the paper. The shortage of the current method was pointed out after analyzing it with the fractal theory. An improved method was proposed then. An experiment result, by comparative analysis on several texture features in a SPOT image, was given. It showed that a more precision result could be gotten by the improved method and therefore it would be helpful for classification.