A Generative Adversarial Network technique for high-quality super-resolution reconstruction of cardiac magnetic resonance images

•DnSRGAN method is proposed for high-quality super-resolution of noisy CMR images.•Feed-forward convolutional neural network is used to pre-denoise the CMR image.•Applied gradient penalty (GP) method solves the problem of the gradient disappearing.•WGAN loss function monitors GAN gradient descent to achieve more stable training.