-1

To control color balance in digital camera exposures it would potentially be useful to be able to change the sensitivity of the sensor on a per channel basis, or alternatively to be able to apply an "exposure compensation" on a per channel basis (where Red, Green, and Blue are the three channels). Is there any camera that supports this ability?

ADDENDUM

Applying color-specific gain in analog circuitry is sometimes called selective gain. Apparently CMOS sensors, because of their design can employ selective gain, whereas CCDs cannot. A reference describing selective gain can be found here:

https://www.extremetech.com/electronics/49052-anatomy-of-a-digital-camera-image-sensors/6

A quote from this source:

In addition to amplification within the pixel site, amplifying circuitry may be placed elsewhere along the CMOS signal chain. This provides different, multiple gain stages throughout the sensor. Amplifiers can apply global gain to increase sensitivity in low light situations. Or selective gain can be applied to a specific color to assist in white balance algorithms or artistic effects.

  • 1
    This question does not show research effort. – user50888 Aug 18 '17 at 15:28
  • Yes there are such cameras. They use three sensors and are designed for video. Video is rarely on topic here. – user50888 Aug 19 '17 at 13:06
2

This is not possible, for the simple reason that a digital camera sensor does not record a colour image. It records a filtered black-and-white image (see RAW files store 3 colors per pixel, or only one?) which is then turned into a colour image via specific calculations (de-mosaicing). There's no 1-on-1 relation in regards to colour between pixels on the sensor and in the image.

But a way to influence the balance of colours in the image is setting a specific white balance (What is the meaning of "white balance"?). (which, of course, only affects the (JPEG) image after exposure.)

An old-style technique would be to use coloured filters in front of the lens. But the utility of this is dubitable with the ease of changing colour in digital post processing.

  • 1
    I am asking about CHANNELS. In digital cameras there are 3 channels: red, blue and green. These channels correspond to pixels on the sensors. Some pixels are filtered red, some green, some blue. The output from each type is directed to a separate color channel. I want to know if there are any cameras that allow ISO senstivity to be changed on a per channel basis. – Clickety Ricket Aug 18 '17 at 21:53
  • 1
    No there are no color channels in the camera. as i explained, and the linked qa shows, there is only a b/w image prior to demosaicing. – ths Aug 18 '17 at 22:05
  • 1
    @ths I dunno, man. There really are different pixels filtered different ways, and if the amplification is set per pixel (as in CMOS sensors) there's no reason in theory a camera couldn't set all red-filtered sensels (whether you want to call that a "red channel" or not — something I think is more a philosophical distinction than a helpful one) to be less sensitive to avoid blowing out reds or whatever. One can even imagine preliminary white balance implemented in this way. – mattdm Aug 18 '17 at 23:41
  • For every pixel there is one sensel of the same type. The filter array simply blocks two "channels" (as you're calling them) per pixel, but these do still have overlap - they're not purely red or green or blue, they're "mostly" reddish, greenish or blueish to put it loosely. – StephenG Aug 19 '17 at 2:10
  • But the filter array doesn't 'block' any color. It just reduces the transmittance of certain wavelengths of light more than others. Some light of all colors still gets through each of the three filter colors. Just look at the response curves for any given sensor. – Michael C Aug 19 '17 at 6:36
1

To control color balance in digital camera exposures it would potentially be useful to be able to change the sensitivity of the sensor on a per channel basis, or alternatively to be able to apply an "exposure compensation" on a per channel basis (where Red, Green, and Blue are the three channels). Is there any camera that supports this ability?

The only way to do this is to shoot RAW and apply different different processing to different channels.

I'm not aware of any application that lets you specifically do that, so you might have to develop the image with different settings to the three channels and merge them in another applications (e.g. GIMP or PS).

  • I am talking about analog channel amplification, not post-ADC processing. – Clickety Ricket Aug 19 '17 at 0:41
1

Modern CMOS sensors do have selective gain per channel, but this gain control seems to be only available on scientific or industrial imagers, not on photographic cameras. This appears to be a purely marketing/feature list decision of the camera companies. The innate ability of CMOS image sensors is that they do have channel-specific analog gain control.

As an example of this, there is the Mallincam SDI camera marketed to astronomy photographers for use in telescopes. This camera uses a typical Panasonic CMOS sensor which is also used in photographic cameras and is one of the members of the Maicovicon sensor product line, now being used for high-end Olympus cameras such as the OM-D M1. If we examine the Mallincam's feature list we find the following settings:

enter image description here

So from this we can see that the sensor supports a basic level gain category ("Kelvin") which can be fine-tuned by selective gain on the red and blue channels. Currently Olympus (for example) does not allow users to access this setting directly, but only gives them a single temperature control. The firmware then adjusts the R and B gain to accomplish this. Other cameras such as Nikon, Canon and Fuji appear to follow exactly the same practice. Note that you do not need control over the Green channel, only the ability to set R/B relative to green. That is why there is no G gain control.

To get full channel-specific gain control currently (Aug 2017) apparently the only options are to either use a scientific imager (such as the Mallincam) or for most cameras to modify the camera firmware and add the feature using binary code, which would not be easy to do. There are some high end Sony and Fuji cameras that do have an option to set the red-blue gain. However, Canon, Nikon and Olympus do not support this option.

  • Many digital cameras provide user control of white balance. Kelvins are the units in which color temperature is measured. – user50888 Aug 19 '17 at 13:53
  • @benrudgers The question is about channel specific gain. Maybe YOU need to do more "research" so you understand the question instead of just assuming my question is wrong somehow because you didn't understand it. – Clickety Ricket Aug 19 '17 at 14:15
  • Setting the white balance from 1700K to 17000K equates to applying gain to a blue channel. Going from 17000K to 17000K equates to applying gain to a red channel. Setting the white balance between them equates to applying more or less gain to one or the other (or in some cases the same amount to both and in others no gain to either). The number of photons collected by the sensor is the number of photons collected by the sensor. When and where the amplification of that number occurs does not change that. – user50888 Aug 19 '17 at 16:10
  • @benrudgers Color temperature is not a substitute for selective channel gain. It is a one-dimensional quantity. R/B channel gain is a two-dimensional balance that can adjust for color casts that are impossible to correct with just a Kelvin curve. – Clickety Ricket Aug 19 '17 at 17:28
  • @TylerDurden CT is one dimensional along a single axis, but WB is 2D within the entire color space of the output format, and that is what the gain applied after digital conversion accomplishes. – Michael C Aug 20 '17 at 4:31
1

Digital imaging sensors in the vast majority of cameras used for creative photography do not measure color information. They measure a brightness value of the light that strikes each pixel.

It is true that for a Bayer masked image sensor the pixels have a red, a green, or a blue filter in front of them. But each pixel still measures light of all wavelengths that make it through the filter and strikes that pixel. Each pixel measures all photons that strike it, regardless of wavelength, indiscriminately.

The green filter allows more light in wavelengths centered on green to pass through than it allows light in wavelengths closer to blue or red to pass through. But a good bit of red and some blue light will make it through the green filter and be included in the total brightness value counted by that "green" pixel.

The red filter allows more light in wavelengths centered on red to pass through than it allows light closer to green or blue wavelengths to pass through. But a good bit of green and a little bit of blue light will make it through the red filter and be included in the total brightness value counted by that "red" pixel.

The blue filter allows more light in wavelengths centered on blue to pass through than it allows light closer to green or red wavelengths to pass through. But a good bit of green and even a little bit of red light will make it through and be included in the total brightness value counted by that "blue" pixel.

Color response curves of a typical CMOS image sensor:
enter image description here

One way to understand this is to remember what happens when we put a red filter in front of the lens when using black & white film. The resulting image is still black and white, but the colors in the scene that are closer to red are as bright a shade of grey in an image with a red filter as they are in an image taken without the red filter. Not so with colors in the scene that are closer to green or blue. The green items are a darker shade of grey than they were in the unfiltered image and the blue items are a LOT darker shade of grey than they were in the unfiltered image.

But the green and blue items are still visible in the red filtered image at reduced brightness values.

That is how the filters over each pixel on a Bayer masked sensor work - some of all wavelengths of light pass through each one. Only when the amount of light allowed through for adjacent pixels filtered for red, green, and blue are compared can color information be derived from the information collected by the monochromatic sensor. This is the process we refer to as demosaicing or as debayerization.

The reason this works so well is because Bayer masked filters emulate the way the human retina works: We have three sizes of cones in our retinas that are sensitive to various wavelengths of light at different attenuations. But each sized cone is sensitive, to one degree or another, of overlapping wavelengths of visible light. Only when our brain compares the differences between the brightness of each set of cones does it create the colors that we perceive.

Color response of the three sizes of cones in the human retina:
enter image description here

Notice how closely the response of our "red" and "green" cones overlap! But the difference is enough for our brains to differentiate various colors based on the comparative difference between the signal it receives from each set of cones.

To control color balance in digital camera exposures it would potentially be useful to be able to change the sensitivity of the sensor on a per channel basis, or alternatively to be able to apply an "exposure compensation" on a per channel basis (where Red, Green, and Blue are the three channels). Is there any camera that supports this ability?

Pretty much every single color camera with a Bayer masked filter (or the external application used to decode the raw data later outside the camera) applies exposure compensation on a per channel basis. That is what we call "White Balance." But for cameras intended for the purpose of creative photography, which is the application for which this group is concerned, they all do it after the signal from the sensor is converted to digital information.

Why? Because to do it by differing amounts to each set of pixels filtered by the three different filter colors before analog-to-digital conversion without reducing the dynamic range and noise performance of the pixels from each set by differing amounts would require altering the exact colors of each set of the filters in the Bayer mask. If we do it electronically to the analog voltages output by each sensel (pixel well), we would introduce a different amount of 'dark current' noise to the value of each set of sensels filtered with each color and we would reduce the dynamic range of the sensor in a way that could lead to banding in areas of color gradation within the image. By amplifying the signal from each sensel equally (remember, at this point each sensel only has a specific brightness value with no color information whatsoever), we are able to exploit the maximum dynamic range and minimize the read noise created by every pixel on the sensor for a specific ISO setting. These disadvantages may be acceptable for other forms of imaging, such as machine vision and other technical applications. For creative photography, the disadvantages seem to outweigh the benefits.

  • I see your logic, however, I have read specific technical reports that describe something called "selective gain" which is color-specific amplification applied pre-ADC. I have added a link to an article describing it. My question is not whether selective gain exists, it is whether there are any cameras that allow it to be controlled by the photographer, or barring that, whether it can controlled by firmware, or is it just baked into the circuit? – Clickety Ricket Aug 19 '17 at 11:51
  • I did some further research and posted an answer below. You might be interested to know that in the process of doing this I found out that the Maicovicon sensors also have a secret monochrome mode that is also not exposed by cameras either. Current cameras only do post-ADC black and white to JPEG. This is significant because if photographers could use this monochrome mode, they could potentially get improved tonality compared to what is possible in a color mode. – Clickety Ricket Aug 19 '17 at 13:25
  • If one reads the answer carefully one will see that it does not say such a method is not possible. Rather, it gives reasons why it is not implemented in cameras intended for the purpose of creative photography, rather than more technical or scientific applications (such as machine vision, etc.) For creative photography, the disadvantages seem to outweigh the benefits. – Michael C Aug 20 '17 at 4:28
  • Answer has been edited to make this more clear. – Michael C Aug 20 '17 at 4:36
  • As in other similar answers... I really think you're attaching more importance to demosaicing than warranted. Demosaicing gives better results, but if you simply treat the separately filtered sensels as pixels that works too. – mattdm Aug 20 '17 at 16:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.