Image watermarks have been widely adopted to ensure the security of important digital assets, particularly within information systems that involve document filing and processing. Although image watermarks can trace malicious users using screen-shooting, they still suffer from external theft, especially screen-filming (e.g., using a camera or a smart phone to take a picture of the screen). The later cases pose a critical challenge to the security of intelligent information systems. This paper discusses a practical way towards establishing a screen-filming resistant watermarking system after analyzing the defects of existing methods and demonstrating the construction of a screen-level image watermark system that can be deployed in the field. This system can provide real-time image watermarking with only a limited consumption in the terminals’ computing units, and achieve the optimal robustness against screen-filming.