HTML5 is the future. Flash and Silverlight are leaving the market, clearing the way for HTML5 video. One big problem remains: protecting the content. Before you know it, the movie is on the free market. Could OpenCL video watermarking be the answer? DRM is only a temporary solution, and it often frustrates users who just want to watch the movie wherever they want.
Watermarking instead of DRM
Look at e-books, and you see a much better way to keep PDFs from spreading all over the web: personalization. You can do the same with images and videos. The example on top of this article has a very obvious, clearly visible watermark (source). However, many methods are hard to see, so people who want to clean the file easily miss them. That is a clear advantage over DRM, where it is obvious what to remove.
Watermarks also give buyers freedom of use. The only disadvantage: you cannot transfer ownership of a personalized video.
Why personalization works
Awareness is the most important part of copy protection. People have to know the media is watermarked. Then consumers won’t feel tricked when some organization knocks on their door. They would have acted more responsibly if they had known their name was printed on the media.
In the previous century, we bought the box around the DVD, while the content of the DVD belonged to everybody. An actor’s signature belonged on the box. This century, we don’t get a box anymore. We only know we get sued when we share the movie with friends. Looking at the past 20 years, I cannot say that people have a good feeling of ownership.
I believe visual personalization, as shown below, gives buyers a better feeling of ownership than a one-minute warning before the movie starts. You could even add a personal message like “for your 2011 birthday, hugs from auntie”.

How robust is video watermarking?
This 2007 master thesis by Martin Zlomek discusses various ways of watermarking and how attackers can break them. It is quite honest about which methods work and which don’t; see the chapter on robustness.
Re-compression, scaling, cropping, adding or removing noise, blurring and sharpening can all distort the watermark, but they damage the movie too. One attack keeps working well: combining several copies of a movie and blurring out the differences. So once professionals focus on a movie, watermarking is as weak as DRM. The difference: attackers can never tell where else the watermark might hide. They must gamble that they removed everything before they release the movie.
Personalised ads in streaming
When streaming a movie, a rental company could use the same technique to build a better customer relationship. It can add a personalized advertisement anywhere inside the movie. That could even make free viewing possible, which would disrupt the illegal market.
The only disadvantage is that all of this takes a lot of processing power, which could raise the price per video too much. That is where OpenCL video watermarking comes in.
The fast lane: OpenCL video watermarking on the GPU
With video cards and GPGPU, you can watermark a video much faster than with traditional techniques. I found one GPU implementation by Brunton and Zhao, but many more possibilities exist. You can also prepare most of the compression (motion detection, the actual compression, etc.) before adding the extra data. Finishing the video compression with OpenCL means you can start streaming the movie quickly.
StreamHPC thinks along with you about your problem and gets your product developed. Look around the site to see what else we do, and ask us any question via the contact form.
