摘要
针对不同分辨率的PET和MRI图像,探讨医学图像中功能成像和解剖成像的融合问题。提出一种基于小波变换的多尺度分解对人脑PET和MRI图像的融合算法,选取测量活跃性等级-系数分组-系数合并的融合规则,对多尺度小波变换的图像融合方法和普通像素加权平均融合方法进行仿真,并运用熵和交叉熵两种方法评价仿真结果。结果表明,该方法能有效融合功能信息和解剖信息,避免虚假信息引入,得到最佳分辨率的融合图像,提高医学图像的可信度,对临床诊断和治疗有一定参考价值。
In this paper, the fusion of functional images and anatomical images with different resolution of PET and MRI image were investigated, A fusion algorithm based on wavelet multi-scale decomposition was presented to merge a PET image and a MRI image. We selected the fusion rules including the measuring of activated scale; coefficient grouping; coefficient incorporation to simulate the methods of multi-scale wavelet transform and general average of pixels adding. The results were evaluated by entropy and cross entropy. The result showed that the algorithm was very efficient in the fusion of functional information and the anatomical information and avoided false information. Thus the best resolution was achieved. The proposed algorithm is expected to improve the reliability of medical images and provide reference in clinical application.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2008年第4期521-525,共5页
Chinese Journal of Biomedical Engineering
关键词
医学图像
小波变换
多尺度融合
加权平均
medical image
wavelet transform
multi-scale fusion
weighted average