基于分形的肝脏B超图像的识别方法

A B Scan Liver Image Recognition Classifier Based on the Fractal Dimension Model

本文首先引入离散分形布朗随机场模型,得到肝脏B超图像的分形维数与变差函数之间的关系。分形维数可以通过计算图像的变差函数并用最小二乘线性回归法来得到,它被选为图像的特征值,做为分类依据。实验表明,正常肝脏与癌变肝脏的B超图像表现出不同的灰度值空间变化特征。采用不同的滞后距离方向对最终的图像分类结果有一定的影响,使用四个方向的平均值的分类精度最高。

We first get the relationship between the fractal dimension and the variogram of the liver B scan image using the discrete fractional Brownian random field（DFBRF） model. The fractal dimension can be achieved by calculating the variogram and using quadratic programming, which is used in the classfication. The result suggests that the fraetal dimensions of the two groups are different. Using different directions, the classfication has a little difference, and the classfication accuracy with the average result is the highest.