Editing Red

[Shadowman]

Actually John, the red in the plastic canister looks a little hot to me in spots.

I don't know if this is relevant, but when I need colors bang on, I use a Colorchecker target.

Hi Terry,

Thanks for the suggested device, looks worth having.

[chauncey]

John, I don't understand the red color problems over any other color difficulty with respect to overexposure.Would seem to me that using a XRite would bring WB in line and adjusting
exposure would take care of red problems...what am I not getting here?

[Shadowman]

John, I don't understand the red color problems over any other color difficulty with respect to overexposure.Would seem to me that using a XRite would bring WB in line and adjusting
exposure would take care of red problems...what am I not getting here?

Chauncey,

The difficulty is getting reds correct under different lighting conditions at least for me. The camera never gets it right, so some manual adjustments are needed in-camera and also in post-processing. I have difficulty capturing red flowers in sunlight and satin fabrics with artificial light.

[xpatUSA]

Ted,

Thanks for the information, current editing program only allows RGB editing, unless I get my hands on Photoshop copy again.

I've read that LR does use a wide-gamut RGB color space - "behind the scenes" so to speak. It is called "Melissa RGB" and is Adobe's bastardized version of ProPhoto, used while you're editing the source file. We do have an expert (Andrew Rodney) here who can tell you more about that.

So, perhaps editing in LR is OK and saturated reds can be fixed just by backing off the saturation slider a tad before saving as sRGB?

[Shadowman]

I've read that LR does use a wide-gamut RGB color space - "behind the scenes" so to speak. It is called "Melissa RGB" and is Adobe's bastardized version of ProPhoto, used while you're editing the source file. We do have an expert (Andrew Rodney) here who can tell you more about that.

So, perhaps editing in LR is OK and saturated reds can be fixed just by backing off the saturation slider a tad before saving as sRGB?

Ted,

Thanks for the information.

[george013]

Ted,

Thanks for the information.

I still don't understand how you can use a wider gamut as what your monitor can show.

I remember an article of a school of arts in which perspective was explained. One way to add perspective was to place a red subject on the foreground. It is a special colour.

George

[Inkanyezi]

I think the issue is at least twofold. The saturated reds are not solely an editing issue, neither one of colour space.

Any colour displayed on a computer screen has a tri-colour presentation of the additive colours. Hence red is indeed a colour that can be saturated, and saturated in this case means that the two other colours are not present, or very subdued in relation to the red. Magenta is a colour that does not exist on the computer screen, it is represented metamerically by two colours. But red can indeed be saturated in a digital image presented on the screen.

Colour spaces are graphical representations of the colours that can be printed, but actual colours always are what the screen can display, whether the actual colour falls within that space or not. Hence we may work in a colour space which cannot be shown on the screen, but will be distorted, also with the best possible calibration of both camera and computer screen.

A clipped colour cannot be brought back by decreasing saturation. If it is clipped, desaturating will be the same as "adding grey", which is in fact decreasing its red colour but adding the two other colours, green and blue. There will still be no tonality in any area where pixels are clipped, but only a less red tone that is completely flat. By instead of desaturating only bringing down the red channel will do something slightly different. Still no structure or tonality will be discernible in the red, but any green or blue colour present in the same area of the image will be retained, which in some cases may display discernible detail.

We have been into the red problem in previous threads and I have received flack for it as the subject is a bit hard to understand. It is also much more difficult to grasp, when you think of the light meter or the auto exposure as providing "correct" exposure, thinking that any deviation from it would be over or under exposure.

As red really is a colour present in the digital image, and even the sRGB colour space can render it fairly well, there shouldn't really be any great problems with the red. But our often erroneous perception of "correctly exposed" or "under-exposed" may lead us to think that a correctly exposed image is under-exposed.

The problem is at least twofold. When photographing anything that is intensely red with the setting the meter suggests, red will be brought to clipping, whether there are other colours present over large part of the image area or not. The exposure meter will see the saturated red as one full red and two empty green and blue channels, One "white" plus two "black", hence the saturated red will be regarded as only a third of its actual value. If there is shadow over large part of the rest of the image area, as when taking a red rose in a bush, the problem becomes even worse, as the meter might want to bring the dark green up to brighter values, clipping even more of the red.

An example of this is in this SOOC image of a rose taken with A setting and no compensation, with the histogram of the red channel as shown in Gimp:



We can clearly see, that the red histogram covers the whole digital space, from zero to 255, but that there are few pixels at zero and many at two hundred and fifty-five. The red colour is clipped. Hence the image taken as the camera with evaluative metering pattern wants it, is over-exposed. The point to the right on the curve of the histogram, with number 255 under it is the mouse position at one of the petals, which is not shown in this image. Each and every petal has a fully saturated spot on its highest curved surface, which appears flat in the image - "hot spots".

There is no other way than adjusting exposure to recover any tonality in the red petals of the flower. The image below is taken from the same camera position, but with two full stops less exposure. The jpeg image SOOC is very dark on the green leaves, but by adjusting the curve, they have been brought to more visible levels, which is a better representation of what it looked like. Both these images are jpeg images out of the camera. RAW will not need quite as much exposure for un-clipping the red channel, and from RAW somewhat more tonality will be retrievable and the greens might be less noisy.
But whether RAW or jpeg, less exposure is needed to get the reds right,
and it is not under-exposure, but exposure compensation for the error produced by the meter.

A simpler approach in daylight is shooting manually without the light meter but using the Sunny 16 rule, or using an incident light meter.

[Shadowman]

I still don't understand how you can use a wider gamut as what your monitor can show.

I remember an article of a school of arts in which perspective was explained. One way to add perspective was to place a red subject on the foreground. It is a special colour.

George

George,

Thanks for the suggestion. Found this function inside LR, you can use Soft Proofing to determine if image has colors too vibrant to print. The highlighted areas in the attached photo shows areas of concern, these areas can be desaturated if your profiled printer is shown to be incapable of handling.

[george013]

I think the issue is at least twofold. The saturated reds are not solely an editing issue, neither one of colour space.

Any colour displayed on a computer screen has a tri-colour presentation of the additive colours. Hence red is indeed a colour that can be saturated, and saturated in this case means that the two other colours are not present, or very subdued in relation to the red. Magenta is a colour that does not exist on the computer screen, it is represented metamerically by two colours. But red can indeed be saturated in a digital image presented on the screen.

Colour spaces are graphical representations of the colours that can be printed, but actual colours always are what the screen can display, whether the actual colour falls within that space or not. Hence we may work in a colour space which cannot be shown on the screen, but will be distorted, also with the best possible calibration of both camera and computer screen.

A clipped colour cannot be brought back by decreasing saturation. If it is clipped, desaturating will be the same as "adding grey", which is in fact decreasing its red colour but adding the two other colours, green and blue. There will still be no tonality in any area where pixels are clipped, but only a less red tone that is completely flat. By instead of desaturating only bringing down the red channel will do something slightly different. Still no structure or tonality will be discernible in the red, but any green or blue colour present in the same area of the image will be retained, which in some cases may display discernible detail.

We have been into the red problem in previous threads and I have received flack for it as the subject is a bit hard to understand. It is also much more difficult to grasp, when you think of the light meter or the auto exposure as providing "correct" exposure, thinking that any deviation from it would be over or under exposure.

As red really is a colour present in the digital image, and even the sRGB colour space can render it fairly well, there shouldn't really be any great problems with the red. But our often erroneous perception of "correctly exposed" or "under-exposed" may lead us to think that a correctly exposed image is under-exposed.

The problem is at least twofold. When photographing anything that is intensely red with the setting the meter suggests, red will be brought to clipping, whether there are other colours present over large part of the image area or not. The exposure meter will see the saturated red as one full red and two empty green and blue channels, One "white" plus two "black", hence the saturated red will be regarded as only a third of its actual value. If there is shadow over large part of the rest of the image area, as when taking a red rose in a bush, the problem becomes even worse, as the meter might want to bring the dark green up to brighter values, clipping even more of the red.

An example of this is in this SOOC image of a rose taken with A setting and no compensation, with the histogram of the red channel as shown in Gimp:



We can clearly see, that the red histogram covers the whole digital space, from zero to 255, but that there are few pixels at zero and many at two hundred and fifty-five. The red colour is clipped. Hence the image taken as the camera with evaluative metering pattern wants it, is over-exposed. The point to the right on the curve of the histogram, with number 255 under it is the mouse position at one of the petals, which is not shown in this image. Each and every petal has a fully saturated spot on its highest curved surface, which appears flat in the image - "hot spots".

There is no other way than adjusting exposure to recover any tonality in the red petals of the flower. The image below is taken from the same camera position, but with two full stops less exposure. The jpeg image SOOC is very dark on the green leaves, but by adjusting the curve, they have been brought to more visible levels, which is a better representation of what it looked like. Both these images are jpeg images out of the camera. RAW will not need quite as much exposure for un-clipping the red channel, and from RAW somewhat more tonality will be retrievable and the greens might be less noisy.
But whether RAW or jpeg, less exposure is needed to get the reds right,
and it is not under-exposure, but exposure compensation for the error produced by the meter.

A simpler approach in daylight is shooting manually without the light meter but using the Sunny 16 rule, or using an incident light meter.

If I understand you, a red subject is reflecting less light than the light meter in the camera is calibrated for. So if the red colour is dominant the picture is getting overexposed, or the subject is moved to far to the right in the histogram.

George

[Inkanyezi]

If I understand you, a red subject is reflecting less light than the light meter in the camera is calibrated for. So if the red colour is dominant the picture is getting overexposed, or the subject is moved to far to the right in the histogram.

George

Any single colour, not only red, will have this property.

The meter is calibrated for grey, a grey that supposedly reflects somewhere between twelve and eighteen percent of the light that shines upon it, light of all colours in the spectrum. If the colours of what is in front of the camera are skewed in any way vis a vis this "grey", the meter will still try to make them "grey" overall. This assumption of the light meter will push any saturated colour over the edge, except perhaps green, which usually is not present to such a degree that it poses a problem for the sensor.

Other colours than those of the sensor's three additive colours may also be overly saturated and push one or another channel over the clipping* level. Problematic colours that are not within the sensor's three colours are violet and yellow. Violet may clip both the red and blue channels, and yellow can clip all channels, but more often the red channel than the other two. The most problematic single colour is red, as the meter, supposing that it measures light which is equally strong in red, blue and green, receives only red when the red is saturated, assuming that the other two thirds are black. It is also common that saturated blue tones are clipped, but it is less problematic in for example a blue sky than red that clips over the petals of a flower.

This problem is what led to the invention of the incident light meter and the grey card for evaluating exposure with an electronic device as the Weston Master or Norwood. Taking the reading off a grey card effectively takes colour out of the equation, and measuring incident light does just the same. The Norwood Director or sequels by other manufacurers disregard the reflection from the subject and only gives a value that represents the light falling upon it. Measuring light reflected by the subject from the camera position, as a TTL metering system does, cannot do that.

[george013]

Any single colour, not only red, will have this property.

The meter is calibrated for grey, a grey that supposedly reflects somewhere between twelve and eighteen percent of the light that shines upon it, light of all colours in the spectrum. If the colours of what is in front of the camera are skewed in any way vis a vis this "grey", the meter will still try to make them "grey" overall. This assumption of the light meter will push any saturated colour over the edge, except perhaps green, which usually is not present to such a degree that it poses a problem for the sensor.

Other colours than those of the sensor's three additive colours may also be overly saturated and push one or another channel over the saturation level. Problematic colours that are not within the sensor's three colours are violet and yellow. Violet may clip both the red and blue channels, and yellow can clip all channels, but more often the red channel than the other two. The most problematic single colour is red, as the meter, supposing that it measures light which is equally strong in red, blue and green, receives only red when the red is saturated, assuming that the other two thirds are black. It is also common that saturated blue tones are clipped, but it is less problematic in for example a blue sky than red that clips over the petals of a flower.

This problem is what led to the invention of the incident light meter and the grey card for evaluating exposure with an electronic device as the Weston Master or Norwood. Taking the reading off a grey card effectively takes colour out of the equation, and measuring incident light does just the same. The Norwood Director or sequels by other manufacurers disregard the reflection from the subject and only gives a value that represents the light falling upon it. Measuring light reflected by the subject from the camera position, as a TTL metering system does, cannot do that.

I don't think a lightmeter is color sensitive, so also not grey sensitive. The meter measures light intensities and is probably sensitive for a certain range of wavelength. The 18% neutral grey is something out of the printing world: 1 part of black paint and 1 part of white paint gives a neutral grey that reflects 18% of the light. A lightmeter that's calibrated on 18% reflectance will place that color in the middle of the histogram. A lightmeter that's calibrated on 12% or 20% reflectance will place that same color out of the middle of that histogram.
I did search for some more figures on reflecting of light based on colors. I found this one with 8476 colors and there reflectancy. To many to be useful here but I think it gives an idea. It's an spreadsheet document.
http://www.resene.co.nz/swatches/download_LRV.xls

This is also an interesting document
http://www.xrite.com/documents/Liter...r_Guide_EN.pdf



George

[Inkanyezi]

I don't think a lightmeter is color sensitive, so also not grey sensitive.

I think you misunderstood.

The meter is not colour sensitive, by the assumption that all colours are present. A saturated red colour that will clip the red channel will not be measured as clipping, for the simple reason that the two other colours are absent. The red colour is registered by the light meter, which is "insensitive to colour", as if it were neutral - hence grey. Its luminance however is as high as the red channel reflection from a white surface under the same lighting, which will reflect green and blue as well and give a higher reading. The red surface only reflects red, while the blue and green are absent - black. So when you take a reading of a saturated red colour, the meter registers the red as if it were only a third of its actual red luminance, causing the camera to expose the image three times more than correct exposure.

Exposure metering in the camera is done with tri-colour sensors, just as the image sensor, having red, green and blue sensels. When measuring a saturated red, the green and blue sensels in the meter will not measure any light.

[xpatUSA]

I still don't understand how you can use a wider gamut than what your monitor can show.

George

Then you need to research how most editors work.

For example, I could send you a JPEG with an embedded ProPhoto ICC profile. The image RGB data is using a color space that is much wider than your monitor. But you could still edit that JPEG image data and even save it again as ProPhoto. All the time your editor was "using" ProPhoto color space. Your editor meanwhile was "rendering" the image on your monitor as sRGB because your editor is, or should be, color-aware.

[Manfred M]

I can't find the sources right now, but so far as I ca remember reading, Nikon has been metering individual colour channels for quite some time. Canons latest models have gone this way too.

[george013]

I think you misunderstood.

The meter is not colour sensitive, by the assumption that all colours are present. A saturated red colour that will clip the red channel will not be measured as clipping, for the simple reason that the two other colours are absent. The red colour is registered by the light meter, which is "insensitive to colour", as if it were neutral - hence grey. Its luminance however is as high as the red channel reflection from a white surface under the same lighting, which will reflect green and blue as well and give a higher reading. The red surface only reflects red, while the blue and green are absent - black. So when you take a reading of a saturated red colour, the meter registers the red as if it were only a third of its actual red luminance, causing the camera to expose the image three times more than correct exposure.

Exposure metering in the camera is done with tri-colour sensors, just as the image sensor, having red, green and blue sensels. When measuring a saturated red, the green and blue sensels in the meter will not measure any light.

I don't think I can agree with you.

The lightmeter measures the luminance of the light in cd/m2 and calculates an opropriate shutter and diafragma value for the given ISO.
It doesn't care if the light is coming from a red, green,blue, yellow, black or white subject. But what it does, is trying to place that average value in the middle of the dynamic range of the sensor/film, resulting in the middle of the histogram. Not completely, a lightmeter calibrated on 12% or 20% will place that slightly different.

I don't think the in-camera lightmeter uses color for exposure. It has a RGB sensor but the RGB elements are not used for exposure, more for whitebalans and scene recognition. It was somewhere on the Nikon-site, I can't find it back that quick.

So when you take a reading of a saturated red colour, the meter registers the red as if it were only a third of its actual red luminance, causing the camera to expose the image three times more than correct exposure.

I think this is a mistake in thought.

George

[george013]

http://www.nikonusa.com/en/Learn-And...anced-srs.html

It's somewhere on this site. My pc is to slow at the moment.

George

[Inkanyezi]

I don't think I can agree with you.
/.../
I think this is a mistake in thought.

George

Actually it does not really matter that it is an RGB sensor for measurement, it works almost the same with a usual reflected meter.

If you disagree, please explain why the red will clip, not only when taking a field of poppies, but also when taking a red rose as the one in this image with the setting the meter in the camera suggests. There must certainly be something that causes the meter to deliver an erroneous reading, I would like to know what, if not the absence of two colours.

[george013]

Actually it does not really matter that it is an RGB sensor for measurement, it works almost the same with a usual reflected meter.

If you disagree, please explain why the red will clip, not only when taking a field of poppies, but also when taking a red rose as the one in this image with the setting the meter in the camera suggests. There must certainly be something that causes the meter to deliver an erroneous reading, I would like to know what, if not the absence of two colours.

As said before. If the red subject covers a big part of the image and the reflectance of that red subject is different as where the lightmeter is calibrated for it will get wrong exposed. The same as the black cat/white cat. It's the same story as when using the "zonesystem".

Also notice that the rose is not equal red. There are a lot of shadowparts in. And the green leaves. Light is not equal in that picture.

George

[Inkanyezi]

As said before. If the red subject covers a big part of the image and the reflectance of that red subject is different as where the lightmeter is calibrated for it will get wrong exposed. The same as the black cat/white cat. It's the same story as when using the "zonesystem".

Also notice that the rose is not equal red. There are a lot of shadowparts in. And the green leaves. Light is not equal in that picture.

George

Why do you answer with irrelevant comments? Any image with some kind of interesting image matter is composed of all kinds of tones. It is not relevant to the light meter error.

This is not the black cat in a coal cellar or white cat in snow problem. When the meter is calibrated to see all colours, or in the case of tri-colour sensors only three of them, the absent colours will be assessed as dark, causing the light meter to show a lower value than the correct one for the single remaining colour. That should not be too hard to fathom.

[george013]

The lightmeter doesn't have something as a Bayer array. It just measures the luminace. And the camera might correc something based on the different fields of the lightmeter when using matrix metering.

This is not the black cat in a coal cellar or white cat in snow problem. When the meter is calibrated to see all colours, or in the case of tri-colour sensors only three of them, the absent colours will be assessed as dark, causing the light meter to show a lower value than the correct one for the single remaining colour. That should not be too hard to fathom.

Let's be polite. For this moment this sounds as complete nonsense to me.

George