WCSE 2022 Spring
ISBN: 978-981-18-5852-9 DOI: 10.18178/wcse.2022.04.063

Multi-channel Image Super-Resolution Reconstruction Based on ESRGAN

Pingan Qiao, Jing Li

Abstract— Image super-resolution reconstruction, the task of producing high quality images from existing low quality images, is to solve the problems of poor perception effect and unclear texture of image superresolution results. Then a method of multi-channel image super-resolution under ESRGAN is proposed. Firstly, 2D-3D CNN is used to reconstruct the multi-channel image data and is set to collect more image texture information. Secondly, combined with DPN, the generation network is improved to reduce the memory usage of parameters and alleviate the problem of gradient disappearance. Finally, the texture loss function is introduced to ensure the high frequency details of the generated image while ensuring the integrity of the contents. The experiment indicates that compared with the contrast algorithm, the average value of PSNR index increased by 2.31dB,and the SSIM index of structural similarity increased by 0.071. The visual result is also significantly enhanced by the method of multi-channel image super-resolution.

Index Terms— ESRGAN, Image super-resolution reconstruction, DPN, Multichannel image

Pingan Qiao
School of Computer Science and Technology, Xi'an University of Posts and Telecommunications
Jing Li
School of Computer Science and Technology, Xi'an University of Posts and Telecommunications

[Download]


Cite: Pingan Qiao, Jing Li, " Multi-channel Image Super-Resolution Reconstruction Based on ESRGAN, " WCSE 2022 Spring Event: 2022 9th International Conference on Industrial Engineering and Applications, pp. 534-540, Sanya, China, April 15-18, 2022.