基于小波系数邻域特征的CT与MR图像融合方法

CT and MR Image Fusion Based on Neighborhood Features of Wavelet Coefficients

提出一种基于小波变换的像素级CT,MR医学图像融合方法,利用离散小波变换分别将两幅源图像进行多尺度分解,再用不同的小波系数邻域特征指导高频分量和低频分量的小波系数的融合,低频分量采用邻域方差指导,高频分量采用邻域能量指导,最后根据融合图像的各小波系数重构融合图像.实验表明:不论从主观感受,还是采用信息熵和平均梯度两项指标作为客观定量评价标准,该方法都优于传统的融合方法,获得的融合图像有效地综合了CT与MR图像信息,能够同时清晰地显示脑部骨组织和软组织.

A pixel-level image fusion algorithm based on wavelet transform for merging CT images and MR images is proposed. In the fusion scheme, the two source images are first decomposed into several components with different resolutions and directions using the discrete wavelet transform. Then the wavelet coefficients of the fused image can be obtained by using region-based wavelet coefficients. The base frequency coefficient is based on the neighboring region variance, and the high frequency coefficient is based on the neighboring region energy. Finally, the fused image is reconstructed by performing inverse discrete wavelet transform. Experimental results show that the method outperforms the conventional combining method from both subjective feeling and objective evaluation by using the parameters such as entropy and average grads. The optimum fused images synthesize the information of CT and MR, especially in the display of bones and soft tissues.