I can’t speak to these people’s patented algorithm. But motion (http://www.lavrsen.dk/foswiki/bin/view/Motion/MotionTechnology) (apt-get install motion) has been around for more than 5 years and produces the same results.

looking at the results on some video feeds, it clearly does not account for any camera movement as it’s simply just processing video and comparing changes in pixels, but that being said i have a feeling their algorithm is more complex than just looping through the pixels and checking each one for a change as that technique has been around for years. It is probably doing something based on regions and tunneling down to each pixel change.

Patented algorithm? Since when can an algorithm be patented? This is retarded …

This is still dam cool. I will probably incorporate this in my security system im developing. One thing at a time though , need the rest working.

Mike, this is just a crude attempt for this patent-trolling prof to get more publicity for his invention. It goes against the spirit of academic openness *and* against the spirit of hacking.

This isn’t worthy of hackaday.

CHDK doesn’t modify the camera firmware (as it runs entirely in memory), why would they be ‘too yellow’?

this CAN’t be patented, there is prior art:
The effectv program for Linux features an effect called “hologram” or something like that, which does exactly the same thing, but instead of butt-ugly green overlay, it overlays a cool retro-scifi hologram effect.

http://effectv.sourceforge.net/

so don’t worry too much about breaking “their” “patent”.

Patents must die.

You know most all videocodec of the last decades have used motiondetection to compress video.
and this is in fact just comparing pixels, in short it’s backwards and I’m going to say retarded to release this now as so novel and clever, the claim to fame they can make is putting it in the canon, but seeing canon cameras also have video compressors on chip it can probably be done much better using its hardware to assist.
We live in a day and age where cheap consumer cameras can freaking detect when you are smiling and this should impress? come on now.

I quickly skimmed over the patent papers, and it seems the algorithm uses an interesting adaptive background subtraction technique. Putting the whole issue of software patents aside, I’m not sure whether this patent is justifiable at all in terms of originality.

Sorry, this is patented? They do realise that Apple have had essentially this idea/software built into photo booth for many years..?

prior art doesn’t automatically invalidate patents. patents like thse don’t protect the outcome, but the way to get there. Which there are many ways of doing.

this algorithm seem to work well, its short, math light and integer based so its good for a lot of basic hardware, thats one of the key differences with other algorithms.

am I the only person who recognizes that Photobooth (in mac os) has been able to do this for years? It’s not that hard. So congrats to you U of L people for re-inventing something that didn’t need reinventing. This would be acceptable to me if they said that they “ported the ability to filter motion from the background” instead of “developed an algorithm”

The principle of “Background subtraction” exists for about 30 years now. Television too…
Why would this mean that you cannot innovate on it anymore?

On the author’s site, there is a sequence for a version of the algorithm that only requires 1 comparison per pixel and 1 byte of memory. To me, this seems to be the absolute bottom line in terms of computational resources. Not surprising then that you can embed it in a digital camera. Nice demo anyway.

I quite like the way that this algorithm keeps objects hilighted even after they have stopped moving. Does anyone know of a more open source algorithm that does that? I’m using AForge at the moment and I’m looking for a faster/better alternative.