Canon patent application takes in-camera HDR to the pixel level
By Donald Melanson 
posted June 23rd 2010 11:11AM
posted June 23rd 2010 11:11AM
There's certainly cameras out there that offer in-camera HDR (or High Dynamic Range), but none that go quite as far as what Canon has detailed in a recently published patent application. The short of it is that Canon's proposed method would alter exposure values at the individual pixel level, which should effectively emulate what's now possible but stitching together multiple photographs taken at different exposure settings -- and, for that matter, even open up some new possibilities since those single-pixel exposure values could be tweaked at will. Of course, there's no indication as to what type of camera the method would be used in, or any evidence that Canon has actually taken it beyond the patent application stage, so we wouldn't recommend getting too excited about the possibilities just yet.
AWESOME!
@Paul Elmy
Canon, you rock. You've never failed me with your EOS 550D.
@Paul Elmy
Totally awesome and if they save this as RAW then incredibly awesome. And if they are able to do all this in a single shutter then finally HDR will be possible of moving content, which would be fully incredibly awesome.
You'd need some seriously fast processing though.
Shall we suspect this won't be on a prosumer model to start with.
@Paul Elmy
Not even HDR, I would think this would work well for tricky exposure situations as well.
I take a lot of pictures to be able to do HDR, if they do this right I may have to look into getting this.
Actually, I would even like to see it be able to do normal dynamic range even compared to what all current digital cameras can do.
This is why I love Canon :)
sorry to post off topic.. but is anyone from engadget live-blogging the google/verizon/motorola/adobe conference that takes place in 30 minutes?
I could of swore they had an invite.. I would figure you guys would have some sort of post about it even though it hasnt started.. (i.e. how its been done for every other conference [most recently google i/o and wwdc])
Hold on, lemme set all 14 million pixels individually... just hold that pose for about 17 years...
If they're not taking multiple exposures with different aperatures and/or exposure time, then it's nothing new that can't be done on the RAW image already in post-processing. Maybe the patent is just to do it in camera? Or is this another example of the US patent system allowing a patent on something that is painfully obvious and/or has been done for years as prior art?
@oldosguy Ok if I had read the source -- they *are* proposing to change the exposure time pixel-by-pixel in a second exposure, after a first exposure establishes the amount of light at each pixel. So it *is* something better than what can be done in post-processing. Still don't think it should be patentable.
@oldosguy Our patent system is pathetic and a mess. It allows stuff that absolutely should not be patentable, to be patented. And then there is the worlds crazy copyright terms now. its Life+70 years almost everywhere.
Fujifilm is already doing something like this with their EXR technology:
http://www.fujifilm.com/products/digital_cameras/exr/features/page_02.html
@Grusic
Was about to say exactly that, and they're on their third model.
Do you think it will be included in firmware upgrades, or only for new cameras with it built in?
@steakman It will most likely require a lot more processing, so I would predict that this would be integrated into higher end cameras with a newer DIGIC chip (DIGIC V, maybe).
@steakman Then again, maybe you might see it for the cameras with dual DIGIC IV chips.
@aschettler Yeah I was gonna say, my 7D should be able to handle it
@steakman
Doesn't CHDK do something like this in existing cameras? I might have to look into that.
I guess if this allows you to 'see' the result right away, that would be an improvement.
This works by using an electronic shutter for each pixel. The mechanical shutter is left open for enough time to collect some light on the darkest areas, so they do not appear completly black. The pixels receiving more light are blocked before that time, and their accumulated charge read out.
This way you collect more photons on the dark parts of the image, and at the same time you avoid saturating the pixels on the bright parts. Human eyes have much higher dynamic range than digital cameras, that's why only some pictures with uniform ilumination and materials look real. And that's why a combining 3 or 5 pictures taken with different exposures gives a more convincing result.
Fujifilm does this with the EXR sensor, half the pixels shot one exposure, the other half shot that exposure -1 EV, or -2 EV: Then the software conbines both. This is done in a single shot, so yes, HDR of objects in movement is possible since a year ago.
Its been done already. I design systems that use cameras that have the same capability. They have been around for at least 5 years.
Man am i not a fan of HDR, sensors over time will get better at dynamic range putting this HDR post processing bs to a rest finally.
Check out a rant I did on this
http://froknowsphoto.com/hdr-rant-2-and-samples/