In surveillance video transmission, the quality of the video will be greatly affected when transmitted at a low bit rate due to limited bandwidth. To combat this issue and enable quicker transmission without sacrificing image quality, we’ve introduced SVRNet (short for Surveillance Video Restoration Network), a Video Super-Resolution (VSR) model tailored for enhancing downscaled and compressed videos post-transmission. It incorporates a distinct “separate-process-merge strategy” to segregate the foreground and background, which are then adaptively enhanced using different SR model and finally output the merged SR results. Furthermore, we significantly enhance video quality by incorporating a novel GTGE module as a substream architecture, employing high-resolution frames to refine the output, all while only requiring a minimal amount of network bandwidth. Extensive experiments demonstrate that our SVRNet and GTGE modules can effectively super-resolve the surveillance videos and outperform other state-of-the-art models.