Nowadays, a surge of several research interested in deep learning has been sparked for restoring historical documents and images, since most of them can be affected by many degradation problems that can rise recognition and processing difficulties. The proposed method of HIR-GAN-DWT composed of two main ideas: generate an augmented dataset using the Generative Adversarial Networks (GAN) and incorporate the Discrete Wavelet Transform (DWT) to boost the quality of the historical images. More specifically, the first objective of this paper is to profit from the power of GAN to generate synthetic data to achieve the best deep learning-based diagnostic performance in a limited dataset. Then, the missing details of predicted images are boosting using DWT. Experimental results and evaluation parameters exhibit the efficiency of the proposed method which can be used efficiently for historical image and image acquisition to obtain enhanced images. Also, the proposed approach is suitable for real-time applications and can be easily adapted to devices that support limited memory with predefined operations.