Forensic Watermarking System Implementation
Budget: $250 – $750 USD
I am in need of a professional who can implement a comprehensive forensic watermarking system into my application for DRM. You will use an open source tool (see example GitHub links below) to embed userIDs into images and videos with spread spectrum marking or something else. You will deliver a watermark embedder within the application to watermark content, an admin page to manage data, and a detector function within the admin page to extract the watermark from images or videos. While there are quality open source watermarking systems for simple images, I am looking for someone to implement several improvements:
1. Video marking/detection
2. Computer vision for human tracking to apply the watermarking specifically to the human figures in images and videos. (see linked paper) (https://digitalcommons.sacredheart.edu/cgi/viewcontent.cgi?article=1100&context=computersci_fac)
3. Intensity-based scale-invariant feature transform embedding and detection to enable detection of photos or recordings taken of a screen. (see linked paper) (http://home.ustc.edu.cn/~zh2991/18TIFS_Screen/Screen-Shooting%20Resilient%20Watermarking.pdf)
4. Encryption of the userIDs using a pseudorandom encryption and reversing it in the detector
5. A detector which can work with screenshots or photos of videos of watermarked videos. In other words it will need to find the right frames in the original video to compare against screenshots or clips.
Delivery are commits into a GitHub repo (embedding system, admin page with data and detecting page)
Examples of open source tools that could be used: https://github.com/syvaidya/openstego, https://github.com/cedricbonhomme/Stegano, https://github.com/DAI-Lab/SteganoGAN
To show me that you are a serious contender, please make some comment about something within the screen shooting resilient watermarking paper linked above.
1. Video marking/detection
2. Computer vision for human tracking to apply the watermarking specifically to the human figures in images and videos. (see linked paper) (https://digitalcommons.sacredheart.edu/cgi/viewcontent.cgi?article=1100&context=computersci_fac)
3. Intensity-based scale-invariant feature transform embedding and detection to enable detection of photos or recordings taken of a screen. (see linked paper) (http://home.ustc.edu.cn/~zh2991/18TIFS_Screen/Screen-Shooting%20Resilient%20Watermarking.pdf)
4. Encryption of the userIDs using a pseudorandom encryption and reversing it in the detector
5. A detector which can work with screenshots or photos of videos of watermarked videos. In other words it will need to find the right frames in the original video to compare against screenshots or clips.
Delivery are commits into a GitHub repo (embedding system, admin page with data and detecting page)
Examples of open source tools that could be used: https://github.com/syvaidya/openstego, https://github.com/cedricbonhomme/Stegano, https://github.com/DAI-Lab/SteganoGAN
To show me that you are a serious contender, please make some comment about something within the screen shooting resilient watermarking paper linked above.