提出了一种基于局部二值模式(LBP)和纹理模式统计进行医学图像检索的方法,计算了LBP和局部方差的联合直方图,改进了Log-likelihood统计距离度量算法。通过仿真表明:改进的Log-likelihood统计算法比Log-likelihood统计算法检索准确率高:与基于Gabor纹理特征图像检索相比较,该局部二值纹理模式检索算法检索准确率能提高8%以上。
This paper presents a novel medical image retrieval approach based on local binary patterns combining with texture patterns statistic. Joint distribution between local binary patterns and local variance is computed and the measurement of Log-likelihood statistic distance is improved. Extesive simulations clearly show the superiority of the improved Log-likelihood statistic. Comparison with Gabor features using medical image dataset indicates that of Gabor features.