New HDR Mode in EOS 5D III

[compupix]

According to Mason Resnick at Adorama (http://www.adorama.com/alc/article/Canon-EOS-5D-Mark-III-First-Look] [url]http://www.adorama.com/alc/article/Canon-EOS-5D-Mark-III-First-Look[/url],) the new Mark III will be able to merge multiple exposures in-camera to create HD images.

This sounds like a nice feature, but what I would rather see is an HDR mode that uses a third of the pixels dedicated to each of the following: lower ISO, nominal ISO, higher ISO. I think the camera has enough pixels to do this. That would dramically increase the camera's dynamic range plus it would make it possible to shoot subjects that move in HDR.

 

[awinphoto]

I guess it would depend if the camera auto shoots 3 consecutive shots or if you physically have to fire 3 shots and let it do it's magic. I'm more worried if any halos will occur or any increased noise as what could happen in computer processed HDR. If it's like the iphone HDR maybe ok for slow moving subjects but not fast moving...

 

[tt]

According to Mason Resnick at Adorama ( [url]http://www.adorama.com/alc/article/Canon-EOS-5D-Mark-III-First-Look]http://www.adorama.com/alc/article/Canon-EOS-5D-Mark-III-First-Look] [url]http://www.adorama.com/alc/article/Canon-EOS-5D-Mark-III-First-Look[/url],) the new Mark III will be able to merge multiple exposures in-camera to create HD images.

This sounds like a nice feature, but what I would rather see is an HDR mode that uses a third of the pixels dedicated to each of the following: lower ISO, nominal ISO, higher ISO. I think the camera has enough pixels to do this. That would dramically increase the camera's dynamic range plus it would make it possible to shoot subjects that move in HDR.

Could you force the camera to ISO bracket? FIngers crossed that the Magic Lantern team can crack open the 5D Mark III - because what you're asking for regarding HDR video is in some ways what they've achieved with video for other models

 

[Terry Rogers]

I've thought about this or a variation of this.

I have no idea if this is possible...but...

Could a sensor be designed in such a way that each photosite could have individual iso's applied to them? To clarify, my understanding is that with current sensors, when you increase the iso, you do so across the entire sensor. However, would it be possible to have the equivalent of iso 100 for pixels just before they become blown out, and simultaneously push the iso up for pixels in the shadow regions to avoid loosing detail in the shadows. This would allow for additional dynamic range, squashing pure highlights before they become blown out and boosting pure shadows so there is detail. Then apply some sort of curve for everything in between so you could fit an extremely high dynamic range image into a single exposure.

Just a thought. I have no clue if it is even possible as far as technology goes. But if such a system could be implemented, I think you could potentially fit a scenes dynamic range of 17 - 18 stops into a 14 bit image (iso 100 for the highlist pixels and iso 3200 to boost shadows so they are not 0,0,0 but actually contain detail. mid tone iso around 400).

Any thoughts if such a sensor would be possible?

 

[Kernuak]

Yes, there is an HDR mode, but nobody in my group for the presentation by CPS yesterday had any interest in it, so I don't know any more details.

 

[bvukich]

Any thoughts if such a sensor would be possible?

For space and complexity reasons, it wouldn't feasible to have a separate op-amp for every pixel which would be required to have pixels that have adjustable sensitivity relative to each other.

However, if you were content with a fixed difference in sensitivity, that can never be turned off I suppose you could (at least theoretically) have the gapless microlens array that covers the sensor made with a couple levels of ND filtering built into it laid out in a grid. But that obviously has it's own problems.

 

[Fleetie]

Surely the sensor output are clocked in banks into an array of a few (or a lot?) of op-amps. You could set alternate op-amps to have low gain or high-gain before reading the sensor out into the banks of amps.

So I don't think you'd need individual amps for all the pixels. That would be insane.

It could be done.

But I doubt it'd really be worth the loss of resolution which you would suffer by effectively going from 22MPix at the same sensitivity to, say, 22/4MPix of composite pixels, each consisting of 4 sub-pixels of varying sensitivity.