Incredible video compression claims are a morbid fascination of mine – usually it’s obvious that the claims are going to go down in flames, and it’s just a matter of watching for the inevitable disaster. I’m using the term incredible in its literal sense, a synonym for unbelievable, which comes in to play when somebody comes out of nowhere to say they can improve the state of the art by a factor of 10:1.
So when I saw this press release from CipherFlux, I couldn’t help having that same feeling of “here we go again”. CipherFlux seems to be saying that they can pull off a roughly a 6:1 improvement over the best quality DivX encoding, a feat that would have to be considered remarkable.
Interestingly, I found that CipherFlux’s founder Jerzy Urbanik, was very happy to talk with me about what his team has been working on. Since this is blog format, and we don’t suffer from space constraints, I’ll go ahead and give you the bulk of our Q&A.
The following is the transcript of an email interview with Jerzy Urbanik, President and founder of CipherFlux:
Mark: Your company seems to specialize in optimized security algorithms. How did you get sidetracked into video coding?
Jerzy: We founded CipherFlux with one premise: we decided to challenge the concerns of speed in the field of cryptology, and at the beginning that was our primary area of research. One day, while having a video conference, one of our engineers said: “Well, it would be nice if we could apply our method to improve the quality of video transmission”. “Why not” I said. That is how it all started.
Most of us have interdiscipilinary knowledge which helped us with finding solutions in areas that are seemingly isolated, video compression and cryptology. Traditionally these fields were developed separately because the problems they examined were very different and the techniques they used appeared unrelated. Recently there is a growing consent among the research community that the two fields are quite closely related.
Mark: You’re claiming that you have higher quality than the best MPEG-4 type decoder (comparing to DivX) with compression ratios four times greater. What’s the secret that allows you to do this? You must be using some new techniques, right?
Jerzy: Yes we are, but at this moment we don’t want to disclose any technical details since our research relates to cryptography and this is a sensitive subject. Our technology implementation into video compression allowed us to achieve greater results and to be sure that nobody approaches compression from this angle we have kept it secret. At this point our method is totally lossless. To prove to ourselves that we are going in the right direction we used a simple 197MB video file and we compressed it down to 90KB. But that is just the beginning. There is still a lot that can be done in the area of lossless compression before we move on to lossy compression. For comparison purposes we compressed the same video using the latest version of DivX (ver. 6.22) which utilizes MPEG4 LOSSY compression. The results were impressive, of course, and to our benefit.
Mark: As far as you know, are you in danger of stepping on anyone’s patents?
Jerzy: Our method/algorithm applied to cryptology is totally unique. We did our homework and as far as we know nobody approaches video compression from that angle.
Mark: If this pans out to be as good as you claim, how do you think you will be able to commercialize it? This is the notoriously tough part of the business, turning a good algorithm into a standard that lots of people use.
Jerzy: This is a very good question. It’s true that turning any algorithm into a standard has always been an issue, especially for a new technology. But we believe that as long as the greatest hurdle for efficient data transmission is bandwidth capacity – that limits speed, size, and quality of any transferred music, voice or video file – the industry will welcome our solution when presented with a tool to overcome those limitations.
Mark: When can we expect to see some demos?
Jerzy: We estimate that demo will be available in 2007.
So there are a couple of red flags here, not surprisingly. The first is that Jerzy takes his press release one step further by saying that his algorithm is completely lossless. To be claiming a completely lossless algorithm that bests DivX best quality by 5:1 is truly incredible, and will definitely require big-time scrutiny.
Second, the press release (as of this writing) says:
Our newest tests and the possibility of independent tests will be avaliable after July 15 2006.
In the interview, Jerzy backs off that and says 2007. I wish I could tell you how many times I’ve heard of an incredible compression technique that just “needed a little bit of tweaking” and never managed to ship.
As always, my final analysis in a case like this is extreme skepticism. But I won’t say that these claims are impossible, just incredible. The acid test for CipherFlux, just like anyone else, is to provide reproducible benchmarks using standard video tests that let us perform apples-to-apples comparisons against state-of-the-art compressors. In general, most incredible claimants never get that far. Jerzy says they will, and until then, we’ll remain skeptical but open-minded.