Hybrid Integration of Transforms with Neural Network based Fusion Techniques for clinical and Healthcare Applications

Abstract

The prime objective of Hybrid Multimodal Medical Image Fusion (HMMIF) method is preservation of important features of images and details about various images from source for creating a visually robust enough single fused image provides a very promising diagnostic tool with numerous clinical and healthcare applications. The Non subsampled shearlet Transform (NSST) with Pulse Coupled Neural Network (PCNN) based hybrid algorithms are proposed for MMIF in this paper. In the proposed method, initially input images are decomposed to less and high frequencies with the application of NSST. The components with lesser frequency are applied with averaging fusion rule. The maximum fusion rule with PCNN is applied on high frequency components. The coefficients produced by every frequency bands are inverse transformed to provide fused images. The proposed algorithms provide the best fused images without distortion and false artefacts. Comparison of proposed technique is done with the pre-existing conventional techniques. The images obtained by fusing both sources’ content with the help of the above algorithm gives the best with respect to visualization and diagnosis of the condition.