基于非分离小波的多模态医学图像融合算法*

针对传统可分离小波医学图像融合过程中存在部分边缘丢失和纹理信息模糊的问题，根据CT/MRI图像内容的互补性，提出整合及突出图像细节的医学图像融合算法。该算法在非分离小波分解框架下，对反映图像近似内容的低频分量采用区域信息熵加权融合规则，对反映图像细节特征的高频分量提出了区域亮度细节占优加权的融合规则；最后通过非分离小波逆变换重构融合图像。实验结果表明，该算法能在保留源图像信息的情况下增强融合图像的细节及亮度信息。

Multi-modality medical image fusion algorithm based on non-separable wavelet

In the view of this situation that in the fusion process of the traditional separable wavelet medical image, there were problems of lost part edges and blurred image texture. Based on the mutual complement of CT/MRI image contents, this paper proposed a fusion algorithm of integration and highlight for the medical image details. Under the non-separable wavelet decomposed frame, used a region information entropy weighted fusion rule for the low-frequency component which reflected approximate content, and used a region brightness details priority weighted fusion rule for the high-frequency component which reflected features and details of image. Finally, reconstructed the fusion image through an inverse transform of non-separable wavelet. Experimental results show that this algorithm not only reserves the source image information, but also enhances the fusion image details and brightness information.