基于小波变换和一致性检验的多模态医学图像融合算法

目的选择已配准后的多聚焦医学图像以及MRI/CT灰度图像为实验素材,以探究不同的融合策略对图像融合效果的影响。方法在对多模态医学图像融合时,低频融合分别采取了加权平均、取极大值法、区域能量以及区域方差的对比实验。高频融合分别采取了区域能量、区域方差以及滤波后基于邻域窗口的一致性检验的对比实验。结果通过对融合图像主观(融合效果)与客观(灰度直方图、边缘提取、性能评价)对比分析,找到了多模态医学图像融合的最优融合策略。结论当低频选择局部区域方差融合,融合后的图像轮廓清晰、边缘较完整;而高频选择滤波后基于邻域窗口的一致性检验,融合后的图像更好地保留和加强了源图像的细节信息。

Multimodality medical image fusion algorithm based on wavelet transform and consistency checking

Objective The image fusion method of wavelet transform has important application value.To explore the influence of different fusion strategies on the effect of image fusion,we choose multi-focus images and standard MRI/CT gray images for source images.Methods For low frequency fusion of multimodal medical image, we adopted the weighted average, the maximum method, region energy and region variance, respectively.For high frequency fusion of multimodal medical image,we adopted the region energy,region variance and neighborhood window filtering with consistency checking,respectively.Results Based on the comparative analysis of fusion image from subjective and objective(gray histogram,edge detection and the performance evaluation)effect, we found the optimal fusion strategy for multimodal medical image fusion.Conclusions The experimental results show that the outline of the fused images is clear and its edge is complete when local variance fusion is selected for low frequency fusion,and that the detail information of source image is reserved and strengthened when consistency checking based on neighborhood window is used for high frequency fusion.