基于NSST域像素相关分析的医学图像融合

Medical image fusion based on pixel correlation analysis in NSST domain

针对像素级多模态医学图像融合信息丢失的问题,提出了一种基于非下采样剪切波变换(NSST)的像素相关性分析(PCAS)的图像融合方法。首先,对源图像进行NSST分解,获得高低频子带。然后,利用提出的中心像素方差计算邻域像素与中心像素的强度相关因子,构建邻域像素相关系数矩阵,并提出将相关性拉普拉斯能量和作为高频方向子带的融合规则。再次,计算低频子带中心像素能量以及邻域像素能量梯度信息,得到低频融合决策图。最后,通过逆变换得到融合结果图像。磁共振图像(MRI)和计算机断层扫描(CT)、单光子发射计算机断层成像(PET)、正电子发射断层成像(SPECT)的脑部图像融合实验结果表明,本文融合方法可以很好地保留源图像的显著信息和纹理细节。

To solve the problem of information loss in pixel-level multimodal medical image fusion,an image fusion method using pixel correlation analysis(PCA)in Non-subsampled Shearlet Transform(NSST)domain is proposed.First,NSST decomposition is performed on the source images to obtain high and low frequency sub-bands.The intensity correlation factor between neighborhood pixels and central pixel is calculated using the proposed center pixel variance,and the correlation coefficient matrix of neighborhood pixels is constructed.The proposed correlation-sum of modified laplacian(C-SML)is used as the fusion rule for high-frequency sub-bands.The energy of the central pixel and the energy gradient information of the neighboring pixels of the low-frequency sub-bands are calculated to obtain the fusion decision map for low-frequency sub-bands.Finally,the fused image is obtained by inverse NSST.The experimental results about magnetic resonance imaging(MRI)and computed tomography(CT),positron emission tomography(PET),single-photon emission computed tomography(SPECT)brain images indicate that the proposed fusion method can well retain the significant information and texture details of the source images.