Face Image Super-Resolution Using Inception Residual Network and GAN Framework

Single Image Super-Resolution (SISR) is an image reconstruction technique that aims to generate a high-resolution image from a low-resolution image. One of the SISR implementations is to reconstruct face images in order to gain more facial information from a low-resolution face images. In this paper, we propose a method to reconstruct face images using a Generative Adversarial Network (GAN) framework that able to generate plausible high-resolution images. Inside the GAN framework, we use inception residual network to improve the generated image quality and stabilize the training. Experimental results demonstrated that our proposed method was able to generate visually pleasant face images with the highest PSNR score of 26.615 and SSIM score of 0.8461.