According to marine accident statistics released in 2023, the number of casualties due to marine accidents has slightly decreased; however, the overall number of marine accidents has steadily increased between 2021 and 2023. To prevent such accidents and take appropriate action, government agencies and local authorities have been promoting safety initiatives and implementing CCTV control systems. However, these CCTV systems face limitations, such as reduced visibility in nighttime environments and difficulty in identifying objects. To address these challenges, this study focused on improving nighttime marine surveillance through the development and application of a low-light enhancement algorithm. The research prioritized the performance of low-light enhancement and real-time image data processing for effective safety monitoring. A comprehensive evaluation was conducted using three different low-light enhancement algorithms and two lightweight deep learning model techniques to identify the optimal solution. The results revealed that applying the TensorRT technique to the Bread algorithm achieved an SSIM (Structural Similarity Index Measure) of 0.7 and a frame rate of 100 FPS. This demonstrates that the combination is the most suitable for both low-light enhancement performance and real-time data processing in nighttime marine environments.