OT - For the "geek" in you

I've been a charter subscriber to Wired magazine, which occasionally has an article that relates to our digital world. The current issue has a piece on a technology called compressed sensing, better described by a post I saw on a different website:

Compressed sensing - a better method of compression?

GK Sullivan , Mar 01, 2010; 10:28 a.m.

I ran across this article today. It concerns itself with compressed sensing and how a mathematical algorithm is used to clarify images in medicine and other disciplines. It mentions the method could well be used in photography, that compressed sensing could easily be used in place of other in-camera compression methods. In my mind, anyway, it could easily increase the dynamic range of the digital photos we take.

http://www.wired.com/magazine/2010/02/ff_algorithm

I'd be interested in any comments from the more geekiness among us.

Maybe i'm not geeky enough to get it, or didnt read enough, but I dont see how that is much different than current compression algorithms. At any rate, 'the better method of compression' part threw me off, cause at least on the web, what is needed is a better method of downscaling documents, not upscaling which is what this article is primarily talking about. It also depends on the application--if all you want is a pretty picture, fine, fill in the blanks between the pixels as much as you'd like--but if there's something in between the pixles that isn't being captured (say, a mosquito at 300 meters that you really want to be included), all the fancy algorithms in the world aint gonna guess that missing part of the puzzle. We'll see, i guess... Thanks for the link, Rod

What's a meter?

Its a big stick that we geeks use to wack wise*** photographers with, Todd

Interesting. I'm no geek, but this seems to be more about speed/efficiency/cost. None of which bother me with the technology I currently use for my photography. But who's gonna complain about better battery life and potentially improved resolution (real or perceived). I'm probably missing something, but I don't see improved dynamic range coming from a much smaller slice of real data/color/tone. Resolution, sure. I guess it's a magic I'd have to see to believe.

I'd personally rather NOT see less "real" data in RAW files for my photography ... but would prefer to watch the compression/software technology improve to take better advantage of "more data" captures. But again, from my own very limited scope of understanding. I do get the appeal for camera makers and other industries.

It is nice to think about getting an MRI in seconds.

This is way out of my league, but I guess I wasn't thinking about getting less "real" data but what would result if you captured the data from a full frame 12 or 24 mpx sensor like we currently do and then it expanded more or better exponentially than current compression technology, resulting therefore in higher dynamic range.

Interesting read. Unfortunately, you are going to see this in the next version of photoshop as this technology really doesn't apply to the output a digital camera will capture.

I am a geek and proud of it. This procedure takes advantage of a sampling methodology which is a real-world way to collect data at set intervals, with hopes of interpolating the missing data enough to be significant. As of right now our cameras do just that using the Bayer sensor, evaluating nearest neighbor values and assigning a discrete value to each photopoint. This method, by what I perceive, takes advantage of what appears to be on par with a Fourier analysis (more so in waveform domain, but has its place in image processing), which has the capability of smoothing the 'surface' greater than what could be achieved with a fully-functional sensor block - a highly complex model in terms of statistics. As a geographer, I collect samples at given points (if my research dictates this) and attribute the samples, enough to attempt to interpolate the missing data - a general law of spatial analysis states that features nearest each other are more related, and ones further away are less related - even on your digital array. The discussion of a mixed-pixel (mixel) has been evaluated and studied among the ranks of Remote Sensing professionals and academics, and that discourse is, well, highly intriguing. If that technology becomes real, we may revert back to 1 to 2-million pixel sensors that will be the equivalent of a 100-megapixel array.

Yada-yada.... sorry folks, back to reality...!!

hey Chris, im a meteorologist--fourier transforms are used all the time by our global forecast models--and you know how accurate our weather forecasts are

Fourier Transforms!!!! Like Laplace Transforms, they bring back bad memories of my engineering days.

My take on it all: "Garbage in equals garbage out". That's my "geeky" opinion and I'm sticking to it!

I'm speechless, I knew I was out of my league.

You're not out of your league, Rod, you're just in a, ahem, league of your own!