Fortifying Fully Convolutional Generative Adversarial Networks for Image Super-Resolution Using Divergence Measures