3D medical image fusion using dual tree complex wavelet transform

Medical image fusion provides content-enriched image which is suitable for a clinical and diagnosis purpose. In this paper, we present a dual tree complex wavelet transform (DTCWT)-based scheme for fusion of 3D medical images. Approximate shift-invariance property and availability of phase information in DTCWT are useful in the fusion process. The approximate shift-invariance property of DTCWT is important in robust subband fusion and also makes it to avoid loss of important image content at multiple levels. On the other hand, the availability of phase information in complex coefficients of DTCWT is useful in encoding more coherent structures of the fused images. The proposed fusion method consists of two steps: i) re-slicing and co-registration of 3D volumes of input images using the statistical parametric mapping (SPM) tool, and ii) fusion of co-registered sliced images using DTCWT. Quantitative evaluations have been performed using five fusion metrics including Entropy (E), standard deviation (SD), mutual information (MI), feature mutual information (FMI) and image quality index (IQI). The results of the proposed method and its comparison with other methods demonstrate the outperformance of the proposed approach.