Deep learning is a sub-field of machine learning, which inspired by the structure of human brain where it is composed with a lot of neurons that each of them able to perform a simple operation and interact with each other to make a decision. In recent years, deep learning has grown exponentially and have brought revolutionary advances in computer vision-based applications such as image classification, object detection, simultaneous localisation and mapping (SLAM), action and activity recognition, age estimation as well as human pose estimation. Based on the literature review, deep learning techniques applied in the vision-based application such as Convolutional Neural Networks (CNNs), Restricted Boltzmann Machines (RBMs), Autoencoder and Sparse Coding mostly focused on solving the feature extraction and classification problems. However, these techniques do not apply to the image processing applications such as image fusion and image reconstruction that require a quantitative knowledge to solve some problems in such fields. The process of obtaining quantitative knowledge in image fusion and reconstruction is a regression problem. In this case, a deep feedforward neural network is applied to solve the regression problem due to its ability to map complicated nonlinear functions.
