基于分形维数的图像纹理分析

Image Texture Analysis Based on Fractal Dimension

利用分数维能够把图像的空间信息和灰度信息有机结合起来的特性,提出了一种基于分形维数的图像纹理分析方法.为了更准确地描述纹理表面的粗糙度,该方法首先将纹理图像进行6种灰度变换并计算其相应的分形维数,同时采用H lder指数作为描述图像纹理的奇异性特征,然后利用提取的纹理特征对图像进行纹理分割.实验结果表明:采用该方法所分割的图像能很好地体现图像的纹理分布.

Fractal dimension involves both the spatial and gray level distribution information of image. A new image texture analysis method based on fractal dimension was presented. Firstly six grey level transformations were done on original texture image, and then the fractal dimension of the transformed image was computed. In order to describe the singularity of texture, another texture feature named Hoelder exponent of original image was also used. Experiments results showed that image segmentation by this method could reflect the information of image texture distribution.