基于DT—CWT变换统计模型的医学图像纹理检索

针对离散小波变换具有平移变化性和弱方向性的特性,本文提出了一种基于双树复小波变换(DT-CWT)统计模型的医学图像纹理检索方法。该方法首先利用双树复小波变换系数的平移不变和多方向选择特性,建立广义高斯分布的统计模型,然后基于该模型提取图像的特征矢量,最后利用改进的Log-likelihood(ILL)相似性测度算法进行纹理图像检索。实验结果表明,该方法的检索查准率达到了82.8%,相比于传统的Gabor算法和小波算法都有了较大的提高,对今后的纹理图像检索具有重要的理论与实际意义。

Medical Image Texture Retrieval Based on DT-CWT Statistic Model

Presents a novel texture retrieval approach of medical images based on statistic model by Dual-Tree Complex Wavelet Trans- form (DT-CWT) for the shift sensitivity and poor directionality of DWT.Establishes at first the generalized Gaussian distribution statistic model upon the characteristics of the shift invariability and directional selection of DT-CWT coefficients.And then obtains the image features by using this model.Finally,carries out texture image retrieval with Improved Log-likelihood algorithm (ILL).The experimental results show that the veracity of new method is up to 82.8%,compared with Gabor and DWT algorithm,which is valuable for texture image retrieval in the future.