An Intelligent Image Compression Method Based on Generative Adversarial Networks for Satellites
Jinglin Li, Youshi Ye, Bo Liu · 2023
This paper analyzes methods to enhance the real-time transmission of remote sensing satellite and ground linkage, pointing out the limitations of the most commonly used on-board remote sensing image compression method, JPEG, and its high bandwidth requirements for satellite-ground transmission. With limited on-board resources, the application of JPEG is constrained, and image compression technology is a feasible and effective approach to addressing this problem. Given the rapid development of deep learning in recent years, this paper summarizes several neural network-based image compression algorithms, considering the balance between image quality and bitrate after decompression and reconstruction, and proposes the use of a lightweight image compression method based on Generative Adversarial Networks. Compared with JPEG, this method significantly reduces the appearance of blocky artifacts, achieves double the bitrate, and reduces computational resource usage by nearly 40% through neural network lightweight processing.