Thermal images, extensively utilized in various communication applications, are concurrently impacted by noise, contrast, and details. However, existing image restoration methods, designed for RGB domain, exhibit suboptimal effects when applied to thermal domain due to a lack of consideration for the interaction between noise and contrast, consequently resulting in detail losses. In this letter, we propose a two-stage deep network based on this interaction for thermal image enhancemnet. Our network decouples the image restoration task into a denoising stage and a contrast improvement stage for simultaneous denoising and contrast improvement. Detail information is extracted, preserved, and fused in the process of the entire network to avoid the detail losses. Extensive experiments show that our propsoed method outperforms other state-of-the-art approachs in terms of PSNR , SSIM, and visual effect.