The world is full of people making miraculous claims for video compression. For example, the hucksters at Euclid Discoveries have been leading their investors on with tales of incredible video compression for years now. No doubt that somebody, someday, might make a quantum leap in video compression, but for the past 25 years it’s been just hard work with slow and steady progress.

A lot of the people who have made these claims in the past are either crackpots or criminals, so I found myself kind of annoyed when Robert X. Cringely popped up with a miracle compression claim in his June 24, 2004 column. Robert isn’t crazy, and he’s not a criminal. He spends a lot of time talking to smart people and trying to synthesize their ideas into frameworks that he can deal with, and he’s pretty good at it.

In this case, he asserted (without references) that the bandwidth of the optic nerve was in the 100Kbps range – thereby implying that we ought to be able to cook up a compression scheme that uses a model of the human eye to provide full bandwidth video at 100 Kbps. It’s not the first time Bob has wandered down this alluring path, he got hyped up in 2002 about Foveating Codecs. (Seen any deployed recently?)

Anyway, the point of all this Bob background is just to point to this paper in which some Penn researchers assert that the bandwidth of the human optic nerve is around 10 Mbps – a bit different from the Cringely numbers. What’s interesting is this means that present coding schemes that provide decent representation of video are operating in the ballpark of that number, which might mean the whole thing kind of adds up.

It’s easy for people to get caught up in a whirl of excitement about impossible video compression, and it happens all the time – just ask the hundreds of innocents who have busted open their piggy banks to buy shares in Euclid. I’ll close with a quote from Bob showing how compelling the idea can be – compelling enough to make people overlook the basic facts:

Oh, and there’s another little side benefit — the end of blindness.

And it is doable, the algorithms have already been worked out and are running today in Matlab.

That was two years ago, so I guess the algorithms are still stuck in their Matlab jail.