MMIF-Net: Multi model image fusion using deep learning convolutional neural network

Abstract

Image fusion plays the major role in many computer vision applications. However, the conventional image processing methods were failed to perform the fusion operation. Therefore, this work focused on development of multi model image fusion network (MMIF-Net) using deep learning convolutional neural network (DLCNN). Initially, preprocessing operation is carried out using median filter, which removes the different types of noises from source MRI and CT images. Then, pixel specific features were extracted using DLCNN model, which performed the feature specific content-based fusion. Here, the DLCNN is used to extract the probabilities of principal component analysis in each MRI, CT region. Then, the post processing operation is implemented using Gaussian filter, which enhanced the overall texture, spatial, spectral regions of MRI, CT images. The simulation results show that the proposed method resulted in optimal performance than the conventional image fusion methods.