An Improved Face Image Super Resolution Algorithm based on ESCycleGAN

Face super-resolution (SR) reconstruction is one of the most popular algorithms in the field of pattern recognition. It can recover high-resolution (HR) images with low cost and high efficiency, and is widely used in public security, image transmission and other fields. This paper proposes a face SR method based on ESCycleGAN. The main purpose of this study is to improve the CycleGan-based SR reconstruction algorithm to retain more detailed information of HR images. Specifically, this paper uses CycleGAN as the basic framework of the algorithm, and introduces the channel attention mechanism to improve the attention to the high-frequency detail information of the face when designing the reconstructed model and the degraded model. Finally, the feature map of the maximum pooling layer is used to measure the image loss, and the high frequency information of the image is preserved as much as possible. Experiments show that the proposed model can effectively reconstruct high quality face images, and the visual effect is better than the traditional reconstruction algorithm.