基于多重分形的肝脏边缘粗糙度分析*

（目的）针对当前在肝纤维化早期CT影像无法检测的问题，提出了基于多重分形理论的肝脏边缘粗糙度分析方法。（方法）该方法计算了图像像素点的粗粒化Holder指数，并用核估计方法估计出图像与该粗粒化指数相对应的多重分形奇异频谱。（结果）实验结果表明该方法能在肝纤维化早期分辨出肝脏CT影像边缘粗糙度的改变，而且与传统分形维方法相比效果更明显。（结论）因此多重分形频谱分析为肝脏图像的检测提供了一种新的途径。

Analysis of roughness of liver edge based on multifractal theory

To solve the problem that liver fibrosis can not be detected by current CT images at early stage, an analysis of roughness of liver edge based on multifractal theory was proposed. The coarse exponent of the image pixels was cmputed, and then its multifractal spectrum was estimated in the kernel estimation method. Experimental results show that the method can distinguish the change of live edge roughness of CT images at the early stage of liver fibrosis. Compared with the traditional fractal dimension method, moreover, it is more effective.Therefore, the multifractal spectrum analysis for the detection of liver image provides a new way.