Digital sensor with variable contrast?

[stefan1]

Is there such a beast as a digital camera with variable contrast? If

not I wonder how hard it would be to make...

 

Another related topic is that of non-linear sensors which could

provide a way to avoid clipping highlights. Again it is a hardware

issue and possibly of too little interest to be commercially viable,

but wouldn't it be nice to have something that responds a little

more like film with over-exposure?

 

I do understand that it is possible to do these things in post

processing but with loss of quality unless the capture is very deep

and not at all noisy.

[hal_mothershed2]

Not exactly sure what you mean, but many digital cameras offer a lot of flexibility in setting various exposure parameters. My (relatively ancient) Minolta D7 allows modifying (and bracketing!) contrast, color saturation, and exposure compensation, among other things.

[donald_choi]

I think a better and more viable idea is to have a digital camera with the capability of taking an instantaneous (or almost instantaneous) double exposure for each shot with the difference in exposures of about 1.5 to 2 stops. The in-camera processing can then "merge" the two exposures together to find a way to "fill in" the blown out highlights as well as brighten shadows. Or, if shooting in RAW mode, you can do this yourself in Photoshop.

 

I know that people do that now for limited (non-action) applications, however, wouldn't it be nice to be able to do this for all applications?

 

Just my thought....

[steve_chan5]

The sensor's RAW output is what you get - can't tweak any further than the base hardware allows. But that being said, the camera usually has a setting that allows you to vary the contrast (within the total contrast range provided by the RAW output). I've noticed that at different ISO's, the available contrast range of the sensor changes on my Olympus. For example, at ISO 400, highlights get blown out when they'd still have detail at ISO 64.

 

If you took 2 frames in the same amount of time as the normal shutter speed, you have basically doubled the ISO of the sensor, so you'll start to see more noise. I also don't know how the cycle time of the aperture and CCD would effect things. I've thought about this auto-bracket feature as well, and it would be handy - but a bracketed RAW files works well for many situations.

[jkantor]

I think that is the future with redesigned sensors. The easiest way would be to just double up - with a high-sensitivity and low-sensitivity sensor at each pixel location, integrating the ouputs. (The large chips have a lot of wasted space.)

[tom_s.]

John Kantor> I think that is the future with redesigned sensors. The easiest way would be to just double up - with a high-sensitivity and low-sensitivity sensor at each pixel location, integrating the ouputs. (The large chips have a lot of wasted space.)

 

Actually, Fuji did this on some of their P&S digicams, such as the FinePix F700. There's a review of it at: http://www.dpreview.com/reviews/fujifilmf700/

[chip]

Hi Thomas,

 

The double sensor on the Fuji 700 is not a good idea, it adds a lot of noise to the sensor. Also the image from the Fuji are not so good as they couild be IMO. I shoot with an EOS 1D, EOS D60 and S230 and have tested my shooting partner's F700 very recently. We were both disappointed in the image quality of the files given all of the gee-wiz technology Fuji has included to improve image quality!

 

I shoot then save to RAW, 1-2 stops is easy to get out of my EOS 1D with it's wide dynamic range in post capture processing, just process both ways, once for th shadows and once for highlights in CapOne Pro and merge the two layers in PS-CS. The D60 is less capable of pulling out the highlights and shadows than the 1D is but the same trick still works fine.

 

HTH

[michael_bolser]

The Nikon N-100 offers dozens of set and custom "curves" that adjust scene

contrast in camera. They can be designed off line and then uploaded to the

camera's operating system. Here is the link to the "custom curves library" at

'Nikonians'.

 

<http://www.nikonians.org/cgi-bin/dcforum/dcboard.cgi?az=list&forum=DCFor

umID79&conf=DCConfID9>

 

There are several consumer cameras that also offer preset scene contrast.

[neil_baylis]

I've always thought this would be a great idea for film scanners. Since the CCD is an analog capture device, they ought to be able to position an amplifier with programmable gain between the CCD output and the A/D Convertor. If this was combined with multi sampling, you could average out the extra noise you would see at high gain settings. Of course, it may be difficult to make such an amplifier with enough stability to give repeatable results. Maybe some high end scanners have this already, but I don't think any consumer/amateur level scanners do. Some claim to have an analog gain control, but I've been told that it only alters the exposure time or intensity.

[stefan1]

Just to clarify - with post-processing I include in-camera processing. What I'm wondering about is a non-linear cmos/ccd element.

 

One application would be to avoid highlight clipping, but another issue I had in mind is scenes where the contrast is very low and I'd like to spend my 12-16 bits on (say) only two stops of exposure.

 

@Neil: Programmable gain in scanners after the CCD does not really make sense to me. Isn't it the sensor itself that has to be modified if you want more film-like response (sigmoid logarithmic instead of linear), plus the light situation in a scanner is about as well defined as you can hope for. Agreed, if you have very heavy negatives you might benefit from having some extra amplification but that seems more like a rescue mission feature rather than something you need on a day-to-day basis.

[david_kieltyka]

One solution would be to create a sensor with 16-bit or greater tonal resolution, using the extra bits as a buffer against overexposure. You could then use RAW conversion curves that more accurately mimic the highlight response of film. I expect we'll see something along this line eventually.

 

-Dave-

[dave_t]

I've wondered myself. Something on the order of a signal dependent gain at the pixel level -- controllable with a simple dial (say, N-3 to N+3). Something that controls the slope (and maybe shape) of the signal vs log(E) curve coming out of the CCD/CMOS. Something that enables good separation in both high and low brightness ranges before post-processing.

<p>

I suspect though that it may be easier just to increase the bit depth -- right now the high end digital backs use the full 16 bits. 32 bits would probably be more than enough if you keep the highlights under control.

<p>

<i>The Nikon N-100 offers dozens of set and custom "curves" that adjust scene contrast in camera. They can be designed off line and then uploaded to the camera's operating system. Here is the link to the "custom curves library" at 'Nikonians'</i>

<p>

I was under the impression that this is post-raw processing. Is that the case, or can you actually program the gain at pixel level?

[steve_chan5]

On another forum, I think someone opined that 12 bits resolution is easy, 14 bits is challenging but still doable and a true 16 bits is very, very expensive because it is hard to get a high enough end to end SNR in the system. I don't know if this is true or not, just relaying what I remember.

 

I doubt that 32 bits are necessary (or even feasible). The MF backs with 16 bits claim something like 12 stops of exposure latitude - that is as good as B&W film.

[michael_bolser]

In response to Dave above, yes, I was referring to Nikon's post processing

functions and not to the detector dynamic range capacity. Had I been more

perceptive, I would have seen that Stefan was really asking for a variable

dynamic range capability. This is problematic...

 

A silicon photodiode has a fixed dynamic range usually measured in db so it

would be difficult to extract a wider dynamic range (in order to correctly record

normally clipped highlights in a high contrast scene) using only one type of

sensor in a given array. This is a weakness of digital photography and is why

the proper exposure (via proper histogram use) is most important.

 

Fuji offers, in their F700, what they tout as wider dynamic range by employing

two different sized sensors in the F700 hexagonal array. Theoretically, the

smaller diodes will be assigned to a higher portion of the voltage (photon)

spectrum while the regular, larger diodes operate from very low levels to

where the high voltages would otherwise be "clipped" thus the combination of

two sets of detectors would deliver a wider dynamic range. It remain to be

seen if this makes a practical difference.

 

PMA 2004 may yield more on this subject from Fuji as their flagship S-2 is

getting on a bit.

[stefan1]

I think Steve Chan has it right about the current status of sensors. The low (dark) bits are going to be noisy in just about any camera... Yet another possible advantage of having (say) a logarithmic response in the sensor is that one could spread the available bit depth down to those difficult-to-resolve dark areas. As it is now in a linear sensor, you spend half of the bits on the first stop of exposure, half of the remainder on the next stop, and so on. This is assuming that you push the exposure all the way up against the right hand side of the exposure histogram, it just gets worse if you don't. Even if you can reach deep bit-wise, it is important to remember that zones are not given equal weight in the capture.