Multimodal Medical Image Fusion Based on Discrete Wavelet Transform and Genetic Algorithm

Abstract

A scheme of medical image fusion for multimodal MRI images is presented here. Basically, every transform-based image fusion scheme comprises of two main steps first image decomposition and second fusion rule according to which the coefficients of image information extracted from individual images at decomposition level are fused to each. The fusion rule is so selected that when the final merged image is reconstructed by obtaining the inverse transform then it has maximum information inserted from individual images. Also, the spectral quality of fused image must be excellent as well as noise free with least fusion error. The scheme used here is applied on multimodal brain MRI images (Flair, T1, T1C, T2) so that the fused image is rich in relevant information about brain ailments like tumor and lumps in brain. The images used in experiment are taken from famous brain image data base, BRATS 2015 and BRATS 2018. For image decomposition, we are using discrete wavelet transform while for fusion model, we are using optimized Bayesian modal using genetic algorithm. The proposed scheme is compared with only discrete wavelet transform (DWT)-based fusion scheme without optimization. It has observed that the optimized scheme is better in performance as compared to simple DWT scheme.