On Combination of Fuzzy Memberships for Medical Image Fusion using NSST based Fuzzy-PCNN

In this article the proposed Medical Image Fusion (MIF) technique uses combination of fuzzy membership val- ues as input to Pulse Coupled Neural Network (PCNN). We named the used PCNN model as Fuzzy-PCNN as the inputs are fuzzy in nature. The spatially co-registered medical images are multi-modal in nature. After decomposing the source medical images using Non-Subsampled Shearlet Transform (NSST) low frequency subbands (LFSs) are fused using the $\displaystyle \max$-selection rule. To fuse the high frequency subbands (HFSs), fuzzy member- ships using multiple membership functions are generated from a specific local-region of the HFSs’ coefficients. Then an L2- norm based ensembling operation is applied to find out the resultant of them. These resultant fuzzy memberships are used as input of PCNN. Inverse NSST (INSST) is applied to the fused coefficients to get the fused image. Visual and quantitative analysis and comparisons with state-of-the-art MIF techniques show the effectiveness of the proposed scheme.