基于Gabor小波变换的医学图像纹理特征分类

Medical Image Texture Features Classification Based on Gabor Wavelet Transform

Gabor小波变换技术对医学CT图像进行纹理特征分类时,由于图像拍摄角度的变化会造成分类的误差。针对以上问题,在Gabor小波变换的基础上提出一种用于分析旋转不变医学图像的方法。该方法采用旋转规范化,即特征元素的循环移位使规范化后所有的图像都具有相同的主方向。实验结果表明,加入旋转规范化循环算子的Gabor小波变换在医学CT图像纹理特征分类时能够达到较好的精确度。

Gabor wavelet transform lacks in its ability to classify the medical CT image if it＇s rotation invariant image. Aiming at the problem, an approach is presented for rotation invariant medical texture classification based on Gabor wavelet transform. Rotation normalization is achieved by circular shift of the feature elements, so that all images have the same dominant direction. Experimental result shows that Gabor wavelet transform with circular operator of rotation normalization has well precision to classify the medical CT image.