学术论文

      基于Contourlet变换和T混合模型的医学图像融合算法

      A Medical Image Fusion Algorithm Based on Contourlet Transform and T Mixture Models

      摘要:
      医学图像融合已经成为医学图像处理领域的热门研究之一.针对基于高斯混合模型的期望最大值融合算法容易导致局部细节丢失的问题,提出了一种基于Contourlet变换的T混合分布图像融合方法.首先通过GIHS(Generalized Intensity-Hue-Saturation)变换将彩色医学图像从RGB颜色空间变换到GIHS空间,进而通过轮廓波变换(Contourlet)获得高频和低频两个部分;然后采用系数绝对值选大法和基于T分布混合模型期望最大法分别对高频部分和低频部分进行融合;最后利用Contourlet反变换获得新强度,将其和PET图像的其他分量通过GIHS反变换得到融合结果.该方法相比于其他的融合方法,具有信息量丰富、清晰度高等优点.
      Abstract:
      Medical image fusion has become a hot research in the field of medical image processing,among which the most classical method called Gaussian mixture models with Expectation Maximum(EM) fusion,the most classical method may lose the local detail.This paper presents a fusion algorithm which is based on Contourlet transform and T mixture models.Firstly,the RGB color space of source images are converted to the GIHS space through GIHS(Generalized Intensity-Hue-Saturation) transform.Secondly,with the Contourlet transform,the intensity component are decomposed into multi-resolution representations,and then the maximum absolute value of coefficient is applied to fuse the high frequency,EM algorithm is used to estimate the parameters of T mixture models.Lastly,the new intensity is obtained by inverse Contourlet,combining hue and saturation to got final result.Experimental results indicate that the proposed algorithm can obtain the results with more functional,spatial information and obtain a better evaluation than other mainstream algorithms.
      Author: Xu Chunyan Song Yuqing Liu Zhe Bao Xiang
      作者单位: 江苏大学计算机科学与通信工程学院,江苏镇江,212013
      年,卷(期): 2017, 40(1)
      分类号: TP391
      在线出版日期: 2017年6月2日
      基金项目: 国家自然科学基金,镇江市科技计划项目,江苏省自然科学基金