Why are red colours always saturated ???

[yb98]

I have noticed that when I shoot red flowers, the red colour is always oversaturated...why ? Does sensors have problems with the red component ?

http://bellik.free.fr/bee2.jpg

[scottbergerphoto]

It looks like that picture is overexposed. Did you check to see the red part of the histogram? Is is clipped? You can't judge saturation on an improperly exposed subject. You also didn't say what camera and color settings you are using.
Scott

[sdommin]

I have noticed that when I shoot red flowers, the red colour is always oversaturated...why ? Does sensors have problems with the red component ?

Flowers (not just red ones) have a lot of colors in the ultraviolet range - colors that our eyes can't always see well, and that don't always reproduce well in digital cameras.

The purpose of a flower is to attract insects, which carry pollen to reproduce the plant. An insect's "gamut" is different than ours.

[Guillermo Freige]

sRGB colorspace is weak in red gamut, that's why is so easy to overexpose it. It can be solved using AdobeRGB instead. Of course you can convert it again to sRGB after postprocessing and taming the reds to avoid cliping, for web publishing or other non color managed applications for example
If no AdobeRGB is used, you can also resolve it partially underexposing the picture.

[Roger_Cavanagh]

Reds can be tricky. Part of the difficulty is knowing when the red channel is blown as many histograms - like on the 10D - don't display individual channel data, but an average across all 3 channels.

It's something that you have to learn to deal with, but there is some good advice in the previous posts.

Regards,

[iwatkins]

Rule of thumb for me:

If the main subject is red, orange or yellow, in bright sunlight, I underexpose by 1 stop on the 10D.

If the overall image appears underexposed afterwards I can recover it, but 99 times out of a 100 the original capture will not have clipped red channel. I shoot RAW.

If the 10D had and RGB histogram, this would be made a lot easier. Does the 1D MkII ?

Cheers

Ian

[maderito]

sRGB colorspace is weak in red gamut, that's why is so easy to overexpose it. It can be solved using AdobeRGB instead. Of course you can convert it again to sRGB after postprocessing and taming the reds to avoid cliping, for web publishing or other non color managed applications for example
If no AdobeRGB is used, you can also resolve it partially underexposing the picture.
Adobe RGB and sRGB have identical Red and Blue color coordinates in their color gamut triangles. Only the Green coordinate differs. The major gamut differences is along the Blue-Green axis of the gamut, i.e. blues, cyans, and greens, much of which cannot be printed or displayed on a monitor. Relatively speaking, there is very little difference in the Red-Yellow portion of the two gamuts and none along the Red-Blue axis. See http://www.rogercavanagh.com/helpinfo/27_gamuts.htm for a graphic representation of the two gamuts.

Some people make a different argument. There are 16 million possible colors that can be mapped in any color space (255^3). In the larger Adobe RGB space, many of these colors are "wasted" in blue-green colors than can't be printed. Meanwhile, remaining colors in the Red portion of the gamut are stretched out. Thus you get get finer gradations among Reds when editing in the smaller sRGB gamut rather than Adobe RGB. The loss of Reds in the sRGB gamut, however, is relatively trivial. In fact, some prefer to edit red/yellow critical images (e.g. containing skin tones) in sRGB.

Where you tend to have problems with Reds is along the Red-Blue axis - when saturated reds move towards a magenta color. As mentioned, that part of the gamut is identical in Adobe RGB and sRGB.

In sum, you have a wider range of colors to work with in Adobe RGB (especially in the greens and cyans) but finer gradations between related colors in the smaller sRGB space. You choose Adobe RGB or sRGB color spaces for different reasons other than "weakness in the reds."

[slin100]

The 10D's histogram is a luminosity histogram. Luminosity is a weighted average of all 3 channels, consisting of approximately 22% red, 71% green and 7% blue in the sRGB color space.

So far, no one has given a satisfactory explanation of the 10D's difficulty with reds when shooting JPG. The reason it is so easy to clip the red channel is due to the in-camera conversion of the sensor's RAW 12-bit RGB data to the 8-bit RGB values supported in JPG files. One might think that the camera would simply map the maximum 12-bit value, 4095, to 255. This is not the case. The camera clips 12-bit values for the red channel above 1350 to an 8-bit value of 255. OTOH, green and blue values above about 2100 are clipped to 255.

This means that the red channel will end up blown in a JPG file much sooner than the green and blue channels, even though sensor may not have been blown. It also means that you lose around 1 stop in the green and blue channels and around 1 1/2 stops in the red channel of dynamic range when shooting JPG.

If one needs yet another reason to shoot RAW, an image that has a strong red component is it. The extra dynamic range is another.

[slin100]

If the 10D had and RGB histogram, this would be made a lot easier. Does the 1D MkII ?

I don't have a 1D Mk II, but I know the answer is, "Yes."

[yb98]

I did some post-processing and the result seems a little better
even if the obtained red colour is not the same as in the reality :cry:

This is the shot without any adjustement in exposure and saturation
in the srgb space

http://bellik.free.fr/red2.jpg

this one the same but in the adobe rgb space (you have to open it in photoshop to see a very slight difference)

http://bellik.free.fr/red3.jpg

and this one with -0,25 in exposure and -8% in saturation in the srgb space

http://bellik.free.fr/red1.jpg

[maderito]

The 10D's histogram is a luminosity histogram. So far, no one has given a satisfactory explanation of the 10D's difficulty with reds when shooting JPG. The reason it is so easy to clip the red channel is due to the in-camera conversion of the sensor's RAW 12-bit RGB data to the 8-bit RGB values supported in JPG files. One might think that the camera would simply map the maximum 12-bit value, 4095, to 255. This is not the case. The camera clips 12-bit values for the red channel above 1350 to an 8-bit value of 255. OTOH, green and blue values above about 2100 are clipped to 255.

This means that the red channel will end up blown in a JPG file much sooner than the green and blue channels, even though sensor may not have been blown. It also means that you lose around 1 stop in the green and blue channels and around 1 1/2 stops in the red channel of dynamic range when shooting JPG.

If one needs yet another reason to shoot RAW, an image that has a strong red component is it. The extra dynamic range is another.

Maybe your conclusions may be right, but the explanation is off I think. The in-camera clipping occurs because the red filters on the CMOS sensor are stronger. Thus red light arriving at the sensor is weaker. Presumably, these and other factors are accounted for when in-camera processing clips and converts the RAW data to gamma-corrected, 8-bit JPEG data.

Reds tend to get clipped in red objects exposed under sunny conditions. Two reasons come to mind: (1) As noted, you can't always see when the red channel is clipped on the camera historgram (unless you have the 1DMKII). (2) Most images are shot under warmish light, that is the red channel levels are to the right of the blue channel levels on the histogram. Errors in white balance or exposure are more likely to clip the red channel first on the right side of the histogram (and blues on the left side of the historgram). We notice the clipped reds which are, of course, brighter.

[KennyG]

Try taking pictures of blue and violet pansies and compare them side-by-side. You may find the results quite surprising and no, there isn't anything wrong with your camera.

sdommin's explanation is the correct one and some plant colors are very difficult to accurately reproduce. Often the only answer is to experiment in post-processing to get it as close as possible.

I must dig out my pansy shots. It confused me for ages until a botanist friend explained it to me.

[slin100]

Maybe your conclusions may be right, but the explanation is off I think. The in-camera clipping occurs because the red filters on the CMOS sensor are stronger. Thus red light arriving at the sensor is weaker. Presumably, these and other factors are accounted for when in-camera processing clips and converts the RAW data to gamma-corrected, 8-bit JPEG data.

I think we're both right. That the red channel receives less light due to the stronger filter is accounted for by scaling the 12-bit to 8-bit transformation a little more aggressively than the other channels.

The explanation I gave comes from John Sheehy. See his post here. (http://forums.dpreview.com/forums/read.asp?forum=1019&message=7521153)

Reds tend to get clipped in red objects exposed under sunny conditions. Two reasons come to mind: (1) As noted, you can't always see when the red channel is clipped on the camera historgram (unless you have the 1DMKII). (2) Most images are shot under warmish light, that is the red channel levels are to the right of the blue channel levels on the histogram. Errors in white balance or exposure are more likely to clip the red channel first on the right side of the histogram (and blues on the left side of the historgram). We notice the clipped reds which are, of course, brighter.
I agree with both points.

One experiment might be to intentionally set a very warm WB to shift the red channel to the left on the histogram. This should reduce or avoid clipping in the red channel. You can correct the white balance later in Photoshop. OTOH, I think this experiment is only of academic interest. You're better off shooting RAW in this situation.

[Guillermo Freige]

I don't know why exactly this happens, but AdobeRGB allows me to keep more blue and red values in both extremes of the histogram, and to some extent also in green one. This is an example:
This is the final processed picture, converted from RAW to AdobeRGB and then postprocessed to reduce reds until they "fit" in sRGB histogram.
http://gfreige.homelinux.org/DR_6307.jpg

I've converted the same image to a 16-bit sRGB TIFF, and to a 16-bit AdobeRGB TIFF.

The histogram from sRGB one is this one:
http://gfreige.homelinux.org/sRGB.gif

And the one from AdobeRGB one is:
http://gfreige.homelinux.org/AdobeRGB.gif

As you can see, reds and blues are better preserved in the AdobeRGB one.

[maderito]

Your histograms illustrate the points I was making:

The Blues and Greens are significantly clipped (on the left) in sRGB when compared to Adobe RGB. Adobe has a larger space in the Blue-Green (cyan) region.

In comparison, the Red is only slightly clipped in sRGB. It just looks a bit more obvious because of the small blip on the right of the histogram. Overall, the Red channel is very similar in Adobe RGB and sRGB.

Incidentally, you could adjust white balance to "unclip" the red channel highlights in sRGB.

The real test is can you see the difference on your monitor or in prints?

[Guillermo Freige]

The real test is can you see the difference on your monitor or in prints?

Yes. In fact, in sRGB the front leaf is a mass of red. The difference is so noticeable I use it to actually "see" when the colors are in sRGB colorspace when working in AdobeRGB, as my monitor is sRGB calibrated.
Regarding reds, I had a picture taken in RAW with the S50 last year of this same flower. I was unable to do anything good with the picture in sRGB (using BreezeBrowser) until I used PS CS Camera Raw to generate a AdobeRGB TIFF from the same image. I was able to recover a lot of detail from the flower.

This is a link of the original embedded JPEG:
http://gfreige.homelinux.org/CRW_0445.jpg

And this the postprocessed one using AdobeRGB:
http://gfreige.homelinux.org/gallery/photos/small/S50_0445.jpg

[maderito]

I'm losing track of your argument.

The JPEG/sRGB example you show is clearly overexposed, especially in the Red channel. You are right - there's is no detail in the reds.

The RAW processed version of the image is better exposed. The reds are less severely clipped. Presumably you adjusted exposure and or white balance during RAW processing before saving it in Adobe RGB space. You recovered some of the red channel information by RAW processing.

It's also possible that there are differences in the way your camera or your RAW conversion software handles color space conversions. I'm not familiar with your camera.

If after RAW processing (adjusting exposure/white balance) you save 2 copies of the image - one in sRGB and one in Adobe RGB - the final result will be very similar to your eyes.

Incidentally, I always process in Adobe RGB. I'm not arguing which is the best capture or editing space. I'm only trying to straighten out misperceptions about what's similar and different about the two color spaces.

To share examples over the web is not useful unless you convert each to sRGB. Viewing an Adobe RGB image on the web without conversion to sRGB will make it look washed out. That may be confusing - but it's essential to understand. And probably you already do!

[Guillermo Freige]

Ok, probably bad example. Forget about it.
The S50 doesn´t allow exp. compensation during RAW conversion using standard tool (FVU) and PS CS alowed that, so a minimal exp. compensation was done, but the S50 RAW file have almost no info passing the clipping point as Canon DSLRs have, so almost all the gain was in the different colorspace used. WB is the same in both pictures. The picture looks darker because the postprocessing, after trying to recover more highlight information using curves.
My point is, if I choose AdobeRGB during conversion I end with more info in all 3 channels, in both ends, but the one I notice more is in red channel. I don´t know if it´s related to the eye sensitivity or what, but at least to me, it´s a real advantage. As I use 16-bit channels, I'm not worried about the "unused bits".
My final colorspace is sRGB, so I use AdobeRGB just as a bigger color container to have more info during postprocessing, and after all the manipulations I convert all to 8-bit sRGB JPEGs. All the images I've posted were in sRGB.

BTW, and in a related topic discused here too, I noticed a huge improvement in color acuracy using C1 and the custom camera profiles provided in etcetera.cc site. In some colors (as some violet flowers) the difference is astounding. The first picture was processed using a "red corrected" version of that profile (the original red output looked too orange to me). If someone is interested I can post some comparisons in a different thread to not to fill this one with even more pictures :)

[Malaxos1]

When I first got the 300D I shot some red flowers and noticed that there was a seriuos loss of detail compared to everything else in the photo. What I did was turn my saturation level in camera all the way down. I just traded my Rebel for a D60 and made sure that the saturation level was low. It is much easier to control the detail by adjusting saturation in Photoshop...Dean

[Jesper]

If one of the channels is blown out, such as the red channel on your flower, you might still be able to save your photo.

Go here and have a look at the tutorial called "Restore Those Clipped Channels": http://www.thelightsright.com/DigitalDarkroom/Tutorials.htm (this was posted in the Post Processing and Printing forum some time ago).

[drisley]

Jesper, that is an awesome link with some amazing tutorials.
Thanks for that!

If someone is interested I can post some comparisons in a different thread to not to fill this one with even more pictures :)

Yes, please do! :)
I've noticed that the default profiles with C1 do have some problems with saturated reds, and I've often wondered about the other profiles out there.