Why do bright red flowers overexpose?

Started May 3, 2011 | Discussions
labnut Senior Member • Posts: 1,016
Why do bright red flowers overexpose?

A recent thread asked this question.
http://forums.dpreview.com/forums/read.asp?forum=1036&message=38216206

I thought I would start a new thread as the technical reasons deserve a discussion of their own.

Photographers with most makes of digital camera have found that red flowers in sunlight come out over exposed with the red channel saturated.

This has a two fold cause.

First the CMOS sensor in the camera has an extended spectral response extending into the near infrared. See the diagram below which shows strong response in the near infrared above 700 nm..

A near infrared filter is normally fitted to limit the response to 700 nm, corresponding to the human eye. In practice the sensor is still responsive to some light in the near infrared range. See Thom Hogan, Shooting Infrared With Digital Cameras http://bythom.com/infrared.htm .

Secondly many red flowers have very strong spectral reflectivity extending into the near infrared. As an example see the spectral response plot for the poppy below.( http://www.reflectance.co.uk , a database of flower reflectance) See also this paper, FReD: The Floral Reflectance Database: http://precedings.nature.com/documents/1846/version/1/files/npre20081846-1.pdf .

It is the additive combination of strong reflectance extending into the near infrared and the partial sensitivity of the sensor to the near infrared that results in saturated, over exposed red flowers.

Diagram from this Kodak publication: http://www.kodak.com/global/plugins/acrobat/en/business/ISS/supportdocs/ColorCorrectionforImageSensors.pdf .

The way to deal with this problem is to spot meter on the brightly reflecting red object and treat it as if it were a highlight(which it is). Then increase your exposure by about 2 EV. The exact value will depend on the lighting and you will need to do some tests to determine the correction value.

Barry Fitzgerald Forum Pro • Posts: 29,888
Re: Why do bright red flowers overexpose?

Won't disagree with any of that..

Though I notice Pentax are pushing reds a bit too much even in raw files (least to my eye) I had similar problems with my Km5d which also is fairly red saturated..less so with the Sony body I have which is more neutral.

Different makers have varying responses on specific colours and hues this is also a contribution to the problem or can be.

Only time I have not had problems with primary reds is on my Fuji EXR compact (using the DR mode) and film which never blows a colour channel ever Film reds look entirely different to digital reds even with fairly high saturation films.

(unknown member) Veteran Member • Posts: 3,992
Re: Why do bright red flowers overexpose?

Color cards and profiling are your friend

In color sensitive conditions, I would never bother to shoot without a properly calibrated scene. This would involve a "proper" color calibration chart and the proper profiling in ACR to match.

Perfect color channels every time.

PS. this applies to RAW btw.

OP labnut Senior Member • Posts: 1,016
Re: Why do bright red flowers overexpose?

Barry Fitzgerald wrote:

Won't disagree with any of that..

Though I notice Pentax are pushing reds a bit too much even in raw files (least to my eye) I had similar problems with my Km5d which also is fairly red saturated..less so with the Sony body I have which is more neutral.

Different makers have varying responses on specific colours and hues this is also a contribution to the problem or can be.

I suspect that is because the different manufacturers fit different near infrared filters.

Pacerr Senior Member • Posts: 1,672
Re: Why do bright red flowers overexpose?

Barry Fitzgerald wrote:

" Different makers (and users) have varying responses . . . [to IR wavelengths]

Apparently that characteristic wasn't designed to annoy DSLR users . . . it seems the plant/bird/insect relationship prefers it that way for a number of practical reasons related to their "sensors".

There's a rumor that a bee/humming bird consortium is trying to negotiate a consulting contract with sensor manufacturers.

Fortunately, we humans don't (yet) have to deal with integrated olfactory sensors in macro lenses!

H2

OP labnut Senior Member • Posts: 1,016
Re: Why do bright red flowers overexpose?

Pacerr wrote:

Apparently that characteristic wasn't designed to annoy DSLR users . . . it seems the plant/bird/insect relationship prefers it that way for a number of practical reasons related to their "sensors".

There's a rumor that a bee/humming bird consortium is trying to negotiate a consulting contract with sensor manufacturers.

H2

Good point, the UV side look blue/green to the bee while red/near-infrared looks bright red to the bird. That is such clever multipurpose design by the flower to account for different sensor designs in the animal kingdom

Gerry Winterbourne Forum Pro • Posts: 19,063
Re: Why do bright red flowers overexpose?

I'm not disputing what you've said but I think there's another factor that can have a big effect. I'm quoting from a printed magazine (Advanced Photographer Issue 2) so I can't give a link.

Briefly, the luminance histogram is based on a weighted average of the three channels - 59% green, 30% red and 11% blue. When the camera's meter tries to achieve 18% neutral grey it does so on the basis of the weighted average luninance (what else could it do?).

These weightings are based on measurements of many typical scenes (green is high because our eyes are more sensitive to green than other wavelelngths, which is also why the Bayer array has more green pixels). As long as our scene bears some resemblance to this typical weighting all is well. But bring in a lot of red and its intensity is only weighted at 30%, so it doesn't push the luminance values up as much as it should (or, to put it another way, it doesn't push the histogram far enough to the right).

That's where the RGB histograms are useful - they show the true values for each colour rather than the weightings. So if there is a lot of bright red in the scene the red histogram is heavily to the right. This allows you to compensate manually.

Bear in mind that the wavelengths we see as R, G and B overlap: this is shown in your diagrams. Even a solid red sheet will show some green and blue on the histograms. It's this, I think, that allows the camera to get rasonably close most of the time but the variability of overlap depending on the exact scene is what makes it hard when the scene gets too far from average.
--
---

Gerry

First camera 1953, first Pentax 1983, first DSLR 2006
http://www.pbase.com/gerrywinterbourne

OP labnut Senior Member • Posts: 1,016
Re: Why do bright red flowers overexpose?

Good point Gerry. I think your explanation nicely extends mine.

Pacerr Senior Member • Posts: 1,672
Re: Why do bright red flowers overexpose?

labnut wrote:

the UV side look blue/green to the bee while red/near-infrared looks bright red to the bird.

And it's so tempting to assign our color definitions to their sensors too. Maybe they just see it as the equivalent of lasagna or sexy lingerie.

H2

richardplondon
richardplondon Forum Pro • Posts: 11,086
Re: Why do bright red flowers overexpose?

Gerry Winterbourne wrote:

Briefly, the luminance histogram is based on a weighted average of the three channels - 59% green, 30% red and 11% blue. When the camera's meter tries to achieve 18% neutral grey it does so on the basis of the weighted average luninance (what else could it do?).

I think it's also worth remembering that (when Live View is not being used) a dSLR's metering will depend on whatever fixed wavelength senstivity and filtration is engineered into the separate metering sensors, not the properties of the main sensor itself. The histogram only happens after the fact.

If this type of metering is systematically confused by particular hue imbalances, and lacks all ability to discriminate these occasions as distinct from other types of scene, then it cannot be programmed around these problems in the first instance; the only answer is to override it - using other knowledge such as experience, or a chimped review - in anticipation.

RP

 richardplondon's gear list:richardplondon's gear list
Panasonic LX10 Pentax K-5 Sigma 10-20mm F4-5.6 EX DC HSM Pentax smc DA 21mm F3.2 AL Limited Pentax smc DA 70mm F2.4 AL Limited +7 more
Barry Fitzgerald Forum Pro • Posts: 29,888
Re: Why do bright red flowers overexpose?

Pacerr wrote:

labnut wrote:

the UV side look blue/green to the bee while red/near-infrared looks bright red to the bird.

And it's so tempting to assign our color definitions to their sensors too. Maybe they just see it as the equivalent of lasagna or sexy lingerie.

H2

Not being the technical expert but I can def see the differences in colour across makers sometimes subtle sometimes not.

If that's down to sensor designs or in camera processing (even for raw) hard to say.

I bet most of the problems would be gone if we had 3 layer colour sensors

Pentax are def a bit too hot on reds even raw that much is obvious that clearly can contribute to the "red problems" I constantly have to dial reds back in LR3 on raw files so much so I've resorted to the camera profile and saving that as a default with reduced reds.

It's mostly an issue when red is the main subject obviously. I'd like to see a more neutral response myself though the natural jpegs are decent enough..bright is a disaster for reds IMO

Pacerr Senior Member • Posts: 1,672
Re: A standard?

Gerry Winterbourne wrote:

Briefly, the luminance histogram is based on a weighted average of the three channels -

Well said, and once again emphasizing that auto-standards, whether WB, EV or AF, must be considered as only a nominal point of departure to be intelligently calibrated and refined to satisfy discrete uses.

The fact that an OEM-standard isn't necessarily a universal standard may be personally inconvenient, but it's beside the point. Camera makers choose their OEM standards to satisfy their own marketing strategy -- to attract a large population of potential customers that have never even heard of DPR!

Camera makers have given the discriminating user a variety of ways to adjust the chosen "auto-standard" to satisfy specific purposes -- even offering convenient pre-sets to assist the unschooled and lazy among us.

It seems strange to me that the convenience of pre-set adjustments is widely favored -- even demanded -- in post-processing software, and yet it's equally disdained as Scene modes for noobies in the camera despite the fact that both functions may be further "customized" to suit the individual user. What's the 'pro' User Mode but a user defined pre-set -- which in some cases may simply replicate a Scene mode anyway?

H2

math guy
math guy Senior Member • Posts: 2,854
Re: Why do bright red flowers overexpose?

Thanks for this explanation, Gerry. I have wondered about this from time to time but never really investigated it much. Just knowing that the camera weights the three channels differently makes it very easy to understand why it blows out the reds sometimes.

Gerry Winterbourne wrote:

Briefly, the luminance histogram is based on a weighted average of the three channels - 59% green, 30% red and 11% blue. When the camera's meter tries to achieve 18% neutral grey it does so on the basis of the weighted average luninance (what else could it do?).

These weightings are based on measurements of many typical scenes (green is high because our eyes are more sensitive to green than other wavelelngths, which is also why the Bayer array has more green pixels). As long as our scene bears some resemblance to this typical weighting all is well. But bring in a lot of red and its intensity is only weighted at 30%, so it doesn't push the luminance values up as much as it should (or, to put it another way, it doesn't push the histogram far enough to the right).

That's where the RGB histograms are useful - they show the true values for each colour rather than the weightings. So if there is a lot of bright red in the scene the red histogram is heavily to the right. This allows you to compensate manually.

Bear in mind that the wavelengths we see as R, G and B overlap: this is shown in your diagrams. Even a solid red sheet will show some green and blue on the histograms. It's this, I think, that allows the camera to get rasonably close most of the time but the variability of overlap depending on the exact scene is what makes it hard when the scene gets too far from average.

-- hide signature --

-- Joe S.
'The laws of nature are but the mathematical thoughts of God.' ~ Euclid

http://www.pentaxphotogallery.com/josephschmitt

 math guy's gear list:math guy's gear list
Sony a6400 Sony E 10-18mm F4 OSS Sony E 20mm F2.8 Sigma 30mm F1.4 (E/EF-M mounts) Sigma 16mm F1.4 DC DN (E/EF-M mounts) +5 more
ernieF Senior Member • Posts: 1,591
Re: Why do bright red flowers overexpose?

Labnut; that's fascinating but probably beyond me. However, the practical implications are well worth knowing, i.e. reds, metering and overexposure and treating brightly lit reds as highlights.

And as you note: I have experienced this phenomena w/ other DSLRs Olympus E1 comes to mind.

Thanks for the techical explanation and practical ways to deal the issue.

Cheers. ernie

 ernieF's gear list:ernieF's gear list
Olympus E-M1 Olympus OM-D E-M10 Olympus M.Zuiko Digital ED 14-150mm 1:4-5.6 Olympus M.Zuiko Digital ED 60mm F2.8 Macro Panasonic 12-35mm F2.8 +5 more
Keyboard shortcuts:
FForum MMy threads