Deep Learning based Compressive Sensing – Radial Basis Functional Neural Network for Image Fusion

Images obtained by different capture techniques of the same scene or objects frequently require fusion to get clear information about the scene. Deep learning plays a vital role in many applications and in handling large volumes of data. Compressive sensing involves fewer samples for image fusion, which addresses some of the issues in traditional techniques but requires more time for reconstruction. This article overcomes the issues of Compressive sensing by employing a trained Radial Basis Functional Network along with deep Convolution layers that reduce the lengthy reconstruction duration and complexity of the reconstruction algorithm. The proposed approach, Compressive sensing Radial Basis Functional Network (CS-RBFN) can reconstruct the images for fusion at a significantly lower cost of computation with improved PSNR and SSIM and the performance is compared with four existing methods.