Ad
Ad
Ad
Pages: « 1 [2] 3 4 »   Bottom of Page
Print
Author Topic: Dual Illuminant Profiles  (Read 35633 times)
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #20 on: May 13, 2012, 07:35:52 AM »
ReplyReply

I read the relevant text on the Handprint web site, and I'm afraid that I am still a bit unclear on the distinctions used when referring to daylight, sunlight, and skylight.  I do understand that daytime SPD measurements vary,
I interpreted "open shade" as follows.  I shot the CC in a location that was shielded from direct sunlight, and only received light from a cloudless mid-day sky.  I tried to minimize exposure to reflected light from other sources, such as foliage.  Is this a proper way to shoot the target for a general purpose single-illuminant profile?

That sounds reasonable to me. Reflected light from other sources would not affect the results if the reflector were spectrally neutral. An ideal setup would be a skylight admitting clear blue sky with the walls painted neutral gray.

Here's an interesting observation: if I load the "open shade" shot of the CC that I made with the D800 and measure the white balance of the lightest gray patch (with the WB eyedropper tool in ACR), I get about 6700K.  If I measure the white patch I get exactly 6500K (and 245/245/245).

The white patch is not spectrally neutral, since it is difficult to make a neutral paint with high reflectance. See Bruce Lindbloom's color checker page. This is why it is recommended to white balance on the patch just to the right of the white patch.

Regards,

Bill
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #21 on: May 13, 2012, 08:09:46 AM »
ReplyReply

The numbers are kind of meaningless in this context once you understand a boat load of colors correlate to any single kelvin value.

See: http://www.ppmag.com/reviews/200512_rodneycm.pdf

See the lines of correlated color temp that run magenta to green? Any color on that line could be specified as 6500K.

If 6500K is meanlingless, you would have a hard time making a profile using this illuminant. Rather than throwing up you hands in failure, you could use light source whose radiance is that of a black body radiator. If the radiator's emissions lie on the Planckian locus, color temperature and CCT (correlated color temperature) will be the same. For practical work, a D65 emulator is often used since daylight varies so much. One approach would be to use a wide spectrum D65 florescent lamp with a high CRI (color rendering index). Hunter labs has a paper giving some useful information on CIE illuminants.

If the source lies on or close to the Planckian locus, white balance in ACR can be achieved with the temperature slider, and the tint slider slider will be at zero.

Regards,

Bill

Logged
digitaldog
Sr. Member
****
Online Online

Posts: 8604



WWW
« Reply #22 on: May 13, 2012, 11:49:07 AM »
ReplyReply

If 6500K is meanlingless, you would have a hard time making a profile using this illuminant.

First off, notice I said in ‘this context’ (that of the OP’s need to understand that an exact definition of the actual illuminant isn’t necessary or even useful/possible).

Quote
Rather than throwing up you hands in failure, you could use light source whose radiance is that of a black body radiator.

The theoretical construct** or something that is real? I’m not throwing up my hands, I’m simply stating that a great deal of this discussion is based on theoretical not real world sources.

Quote
If the radiator's emissions lie on the Planckian locus

Yes if. A big if.

Quote
One approach would be to use a wide spectrum D65 florescent lamp with a high CRI (color rendering index).

IF again you put much credence in CRI.

**http://en.wikipedia.org/wiki/Black_body
Quote
A black body is an idealized physical body that absorbs all incident electromagnetic radiation, regardless of frequency or angle of incidence.

So are we talking about an exact color for the purposes of this discussion with the OP or some theoretical value? I submit the later.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
digitaldog
Sr. Member
****
Online Online

Posts: 8604



WWW
« Reply #23 on: May 13, 2012, 11:53:11 AM »
ReplyReply

Certainly, the ideal method is to shoot it in the same light that falls on the subject, and to build a profile specific for the scene conditions. 

Some would like you to believe that is the case. But is it in practice? I’d like to see proof that is necessary.

Quote
But if I want to build a general purpose camera profile for daytime outdoor conditions, I need to shoot under sunlight.
 

Yes. And again, that is probably all you’ll need.

Quote
And that's where I have been reading conflicting advice.  Some sources advise shooting in direct noon sunlight.  Others recommend "open shade", but do not explain what that means.  I have also read comments that the exact conditions do not matter much.

I’d suggest that open shade and daylight are quite different but aside from those two conflicting examples, Direct noon sun, direct 2pm sun etc, you’ll be fine.

Quote
Here's an interesting observation: if I load the "open shade" shot of the CC that I made with the D800 and measure the white balance of the lightest gray patch (with the WB eyedropper tool in ACR), I get about 6700K.  If I measure the white patch I get exactly 6500K (and 245/245/245).

And there is more to all this than simply building a loading the camera profile. If you load that raw into 3 different raw converters, you get the same reported values? I suspect not.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1147



WWW
« Reply #24 on: May 13, 2012, 04:07:24 PM »
ReplyReply

If you shoot the CC chart in the shade to build your profile from, then the camera sensor's response to the true spectral reflectance characteristics of full spectrum sunlight on each colored patch won't be recorded and measured from. We're talking major hue/saturation and luminance shifts between these two lighting situations. You can see it with your own eyes.

I haven't found one shaded environment that didn't make the gray patches go wonky from the odd color casts associated with bouncing blue light from the sky or greenish/beige from overcast clouded light or tree foliage not to mention the shifts to yellow and blue in hue and/or saturation.

Sunlight is the only light that makes the gray patches look neutral and provides even illumination and hue/sat appearance across all the patches.

You're attempting to measure the camera sensor's response with regard to color gamut and dynamic range capture capability.
« Last Edit: May 13, 2012, 04:11:37 PM by tlooknbill » Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5417


WWW
« Reply #25 on: May 13, 2012, 04:58:28 PM »
ReplyReply

If you shoot the CC chart in the shade to build your profile from, then the camera sensor's response to the true spectral reflectance characteristics of full spectrum sunlight on each colored patch won't be recorded and measured from. We're talking major hue/saturation and luminance shifts between these two lighting situations. You can see it with your own eyes.

Actually, that's not really true...while it's clear that there will be a difference in the white balance, the sensor's spectral response won't be all that different. As long as you are "close" (D50, D55, D65) the resulting profile will work fine. Where's it's important to control the light color when shooting targets is when doing dual-illuminate profiles. The camera's response to tungsten light will be vastly different than pretty much any sort of daylight. Ideally, you should use color accurate tungsten lighting when shooting the target.

For a given camera, a single dual-illuminate profile with daylight and tungsten will be all you really b=need for a camera. The only other profiles you may want to create are profiles for special lighting such as white-light LED (which doesn't really match daylight), fluorescent (accurate daylight bulbs will still have spectral spikes) or mercury vapor–all of which have very different spectral illumination than daylight or tungsten.
Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1147



WWW
« Reply #26 on: May 13, 2012, 07:58:20 PM »
ReplyReply

Quote
...while it's clear that there will be a difference in the white balance, the sensor's spectral response won't be all that different.

I really don't understand how what I see with my eyes which clearly shows a difference in appearance of the CC chart in the shade compared to sunlight translates to what profile building software is deriving from the sensor response data captured from the CC chart under that specific light to correct for in a profile. I have to admit most of the DNG profiles I build of fluorescent and other oddball lights create profiles that don't make much of change over profiles built from full spectrum light source. Using either dual and single illuminant profiles make the biggest difference more so than the type of light shot images are captured under and the profile is applied afterward. See the sample below.

I've never made a profile from a CC chart captured in shade which doesn't seem would provide ideal light levels without higher ISO settings for decent shutter speed to be building a profile from when there's sunlight easily available. Why shoot under shade if you have sunlight?

JulianV indicated:

Quote
I used the CC Passport software to build custom "open shade" profiles for a D700 and D800.  I processed raw files with ACR and PS CS5, viewing on a calibrated NEC PA271W.  On the D700, the results from using the profile were mixed.  In many cases my shots came out with reds, blues and purples that appear over-saturated.  On the D800 the custom profile gave better results - almost indistinguishable from the Adobe Standard profile for that camera.


What's causing the over saturation? It's usually the exact opposite for me where a dual illuminant profile built from tungsten bulb and sunlight CC chart causes over saturation over a single 6500K table profile. Are these just transform errors?
Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5417


WWW
« Reply #27 on: May 13, 2012, 08:32:33 PM »
ReplyReply

I really don't understand how what I see with my eyes which clearly shows a difference in appearance of the CC chart in the shade compared to sunlight translates to what profile building software is deriving from the sensor response data captured from the CC chart under that specific light to correct for in a profile.

First off, the DNG profile ain't being made to profile your eyes...what you see has ZERO to do with profiling your sensor...the sensor has no white adaptation, no ability to adapt to wide dynamic ranges. It's a friggin' sensor with a single set of tri-color RGGB filters over it that dictates a certain response based on a certain illumination. Daylight is daylight. The sun is pretty consistent regardless of the atmospherics that may impact the white balance of the light. Different light sources other than daylight is not so consistent–hence the reason for doing dual-illuminate profiles. Sensors do actually have metameric failure  due to the differences in the way a sensor responds to different  spectral power distributions...daylight is daylight and tungsten is tungsten which is why it's useful to do a dual-illuminate DNG profile.
Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1147



WWW
« Reply #28 on: May 13, 2012, 11:20:45 PM »
ReplyReply

First off, the DNG profile ain't being made to profile your eyes...what you see has ZERO to do with profiling your sensor...the sensor has no white adaptation, no ability to adapt to wide dynamic ranges. It's a friggin' sensor with a single set of tri-color RGGB filters over it that dictates a certain response based on a certain illumination. Daylight is daylight. The sun is pretty consistent regardless of the atmospherics that may impact the white balance of the light. Different light sources other than daylight is not so consistent–hence the reason for doing dual-illuminate profiles. Sensors do actually have metameric failure  due to the differences in the way a sensor responds to different  spectral power distributions...daylight is daylight and tungsten is tungsten which is why it's useful to do a dual-illuminate DNG profile.

We're not just talking about frickin' white balance, bud. We're talking about what characteristics are inherent within an illuminant that seems to have to be profiled for how an RGGB sensor reacts to it.

What's so specific about an illuminant A,B,C,D,E,F,G,H,I,J,K whatever the hell these color scientists are describing that needs to be measured and corrected for in a profile.

You don't seem to be able to make it make sense even in your terse A-hole style response.
Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1147



WWW
« Reply #29 on: May 13, 2012, 11:43:02 PM »
ReplyReply

Here's the single 6500K illuminant profile I made from a spectrally spikey "5500K" Ottlite CFL. Not much difference from the single 6500K illuminant made from full spectrum sunlight posted above.

Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5417


WWW
« Reply #30 on: May 14, 2012, 12:55:16 AM »
ReplyReply

What's so specific about an illuminant A,B,C,D,E,F,G,H,I,J,K whatever the hell these color scientists are describing that needs to be measured and corrected for in a profile.

Standard Illuminate A is about 2856 K...However, the R, G, G, B tri-color Bayer array of a lot (most) of sensors are tuned to "daylight" which has a full spectrum between 400nm (ultraviolet) and 700nm (infrared).

The problem with spectral illumination that has different relative amounts of colored light that is different than daylight is that the tri-color photosite filters vary considerably in efficiency...so while the blue of a tricolor filter may pass enough photons to make a reasonable RGB color image under daylight, that same blue filter may block most if not almost all blue light when shooting under tungsten because first of all, tungsten has a lot of IR (which sensors are more sensitive to) and very little blue light component. This leads to far more noise because of less blue light and a real difficulty in demosaicing because of the blue photosite color errors. So, the spectral response of a sensor will vary considerably based upon the spectral distribution of various colors of light. Which is why it's a good idea to profile under both daylight and tungsten...and any other really odd light source (but not waste your time profiling every potential light source).

Sensors are very sensitive to IR and almost blind to UV (which is why it's stupid to stick a UV filter on a lens for a digital camera). The exact formula for filtering for a tri-color filter separation is a tweaky thing. The designers must design the sensor system for the most reasonable expectation which is daylight. Hence various colors of daylight will be well profiled using a D-50 thru D-65 daylight standard. Whether you have more sun vs. skylight is not an issue for profiling a sensor's daylight response, it's an issue for white balance.

With tungsten which has far more IR and whose red tri-color filter may let in more or too much red/IR light and not nearly enough blue light, you have a totally different sensor response. Hence the metameric failure or inability to profile a sensor under daylight and use it reliably for tungsten.

ACR/LR "tweens" white balance between the Standard Illuminate A (2856K) and D-65. The closer you are to either end of the spectrum, the more the ACR/LR white balance must adjust for the interpolation between the two ends of the white balance scale.

Quote
You don't seem to be able to make it make sense even in your terse A-hole style response.

If I'm not making sense, there is two potential reasons, either I'm not making myself clear or you do not understand the fundamentals...but nothing I've written in this thread is my typical "A-hole" you friggin' idiot (see, you wanna push my buttons, I'll respond) I'm just rying to explain why it's important to do a dual-illuminate profile for daylight and tungsten or other weird light sources which may be spiky and incomplete with their full spectrum of light...

And yes, if you understood the way sensors respond to light you would realize the vast majority of the color corrections issues are white balance, not DNG profiles.
Logged
mac_paolo
Sr. Member
****
Offline Offline

Posts: 419


« Reply #31 on: May 14, 2012, 01:56:52 AM »
ReplyReply

ACR/LR "tweens" white balance between the Standard Illuminate A (2856K) and D-65. The closer you are to either end of the spectrum, the more the ACR/LR white balance must adjust for the interpolation between the two ends of the white balance scale.
...that leads me to a conclusion where Adobe should have really said: take a photo under mid-watt tungsten bulb and another one under daylight.
Everything else is overkill.
Maybe custom DNG profiles for insane artificial lights are the ones where we should put more efforts. Smiley
Logged
julianv
Newbie
*
Offline Offline

Posts: 28


« Reply #32 on: May 14, 2012, 03:26:27 AM »
ReplyReply

Some would like you to believe that is the case. But is it in practice? I’d like to see proof that is necessary.

OK, my statement was obviously too simplistic. Building a profile specific for every scene condition is overkill, except perhaps for those who suffer from OCD (obsessive color-calibration disorder).  Smiley

I'm trying to sift through the heavy tech talk, to come up with a plan for building a general purpose dual-illuminant profile.  So far, I've gathered the following.  Please correct me, if I am going astray.

1) Make one shot of the CC target under daylight.  Shooting in either direct sunlight or open shade are both acceptable.

2) Make another shot of the target under color correct incandescent light.  I'm assuming that a 3500K Solux is OK for this.  Better to use a lower K?

About a year ago, I created a web gallery to compare the variations in color rendition produced by using Adobe's canned profiles, or a custom single-iluminant profile built with CC Passport.  The sample photos were taken with a Nikon D700, and converted with ACR 6.4.  If you visit this page, you can expand your browser window to get the largest image size, and use keyboard right/left arrow keys to switch quickly between images.  To my eyes, the color differences were mostly significant in the blues and reds, which became significantly more saturated with the custom profile. The results were acceptable for these test shots, but were sometimes over-the-top with other images.  Are these results typical?  As noted in my earlier post, the profile that I built for a D800 was much closer to Adobe Standard.
Logged
Nigel Johnson
Full Member
***
Offline Offline

Posts: 124


« Reply #33 on: May 14, 2012, 06:04:57 AM »
ReplyReply

About a year ago, I created a web gallery to compare the variations in color rendition produced by using Adobe's canned profiles, or a custom single-iluminant profile built with CC Passport.  The sample photos were taken with a Nikon D700, and converted with ACR 6.4.  If you visit this page, you can expand your browser window to get the largest image size, and use keyboard right/left arrow keys to switch quickly between images.  To my eyes, the color differences were mostly significant in the blues and reds, which became significantly more saturated with the custom profile. The results were acceptable for these test shots, but were sometimes over-the-top with other images.  Are these results typical?  As noted in my earlier post, the profile that I built for a D800 was much closer to Adobe Standard.

Julian

If I remember correctly the CC Passport software tends to produce more saturated images than the automatic profiles generated by the Adobe DNG Profile Editor (DNG PE) - some people prefer one and some the other. I don't know if you have tried the DNG PE but you can use it on the same profiling images used with the CC Passport software and can generate both single and dual illuminant profiles. The DNG PE is available free from Adobe Labs at http://labs.adobe.com/wiki/index.php/DNG_Profiles.

Nigel

PS The DNG PE is due to be updated to allow editing of the latest DNG profiles, however the current version can still generate profiles that can be used with the latest versions of LR or ACR.
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #34 on: May 14, 2012, 06:58:02 AM »
ReplyReply

Standard Illuminate A is about 2856 K.

...ACR/LR "tweens" white balance between the Standard Illuminate A (2856K) and D-65.

I'm just trying to explain why it's important to do a dual-illuminate profile for daylight...

Jeff,

In the context of photometry, illuminate is a verb. The noun is illuminant. According to the above linked reference, illuminate used as noun refers to a person who is or affects to be specially enlightened.

Just a bit of constructive criticism.

Regards,

Bill
Logged
digitaldog
Sr. Member
****
Online Online

Posts: 8604



WWW
« Reply #35 on: May 14, 2012, 08:29:50 AM »
ReplyReply

Maybe custom DNG profiles for insane artificial lights are the ones where we should put more efforts. Smiley

That has been my experience. I’m sure others will pipe in.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1147



WWW
« Reply #36 on: May 14, 2012, 11:19:03 AM »
ReplyReply

Quote
With tungsten which has far more IR and whose red tri-color filter may let in more or too much red/IR light and not nearly enough blue light, you have a totally different sensor response. Hence the metameric failure or inability to profile a sensor under daylight and use it reliably for tungsten.

So the richer, deeper looking red of the tomato and overall slightly increased saturation using the dual illuminant profile is intended as an attempt to compensate the metameric failure of the dull, orangish looking red caused by the orangy yellow tungsten light?

I'm trying to distinguish the way my camera records the real affects of light on an object from how the software interprets it and whether the unexpected results in the preview is intended or metameric failure and whether it should be fixed in front of the camera or in post for better workflow efficiency. Color science jargon often does not effectively connect the dots in explaining the results we get in our previews.

I have to rely on my observations of how sunlight (not mixed skylight) tends to make objects more vibrant and well defined over any other type of light. And so I have to assume color science is aware of this behavior as well but often find it doesn't make clear when color goes wrong.

You say friggin', I say frickin', no big whoop. Consider both our buttons pushed.

Quote
About a year ago, I created a web gallery to compare the variations in color rendition produced by using Adobe's canned profiles, or a custom single-iluminant profile built with CC Passport.  The sample photos were taken with a Nikon D700, and converted with ACR 6.4.  If you visit this page, you can expand your browser window to get the largest image size, and use keyboard right/left arrow keys to switch quickly between images.  To my eyes, the color differences were mostly significant in the blues and reds, which became significantly more saturated with the custom profile. The results were acceptable for these test shots, but were sometimes over-the-top with other images.  Are these results typical?  As noted in my earlier post, the profile that I built for a D800 was much closer to Adobe Standard.

Most of the differences I see in your parade shot is similar to what I demonstrated in the food shot where I artificially made the gray plate look blue while maintaining warmth in the food which is my understanding of how to demonstrate the effects of color constancy.

Note in your Camera Standard shot the street highlight sheen is greenish while the darker parts are bluish while maintaining R=G=B white while the others tweak this and mostly maintain the same hue/sat appearance in the vibrant primary colors while making the street all neutral or all green. I can see slight tweaks to warm hues in the skin tone of the woman upfront far right.

Which version is derived from the real response of the camera sensor or just software/profile tweaks is anyone's guess. All four look fine.
« Last Edit: May 14, 2012, 11:22:53 AM by tlooknbill » Logged
sandymc
Full Member
***
Offline Offline

Posts: 248


« Reply #37 on: May 14, 2012, 11:57:29 AM »
ReplyReply

This seems to be causing huge confusion - it may be helpful to understand how "dual illuminants" are actually used in processing:

In concept, processing from camera to finished image is in two steps:

Step 1: Convert from sensor readings (aka raw data) to "absolute" color values, in principle in an XY-type space. Critical point is that these are "absolute" values; they depend on the actual color of the light - the wavelength. So in the situation where you shine a tungsten light on a white sheet of paper, what you're measuring is the color of the tungsten light, not that the sheet of paper is white.

Step 2: Convert the absolute (XY) light values to something humans recognize - this is the white balance step, taking into account the fact that the light was tungsten, and adjusting the color of the sheet of paper in the final image to be white, white being what we see.

Dual illuminates are ONLY used in step 1 - they're just a way of better interpreting what the raw data converts to in XY type values. Dual illuminants play no role in step 2. They might give you a slightly better white balance, but that's only because they delivered a better measure of what the light reflected off various parts of the scene was, not because they're playing any role in white balance process.

Put another way, if the sensor in your camera was perfect, dual illuminants would make no difference at all. What they do is allow the software to better estimate what the scene actually looked like from what an imperfect sensor measured.

The reason why Adobe recommends D65 as an end-stop is probably just because that's where color temperatures pretty much cease to make sense as a practical photographic measure. Anything "more blue" is likely not a continuous spectrum light source.

Here's what I'd recommend:

1. For scenes lit by something with a continuous spectrum (daylight, tungsten lights), dual illuminant profiles make good sense, basically because color temperature makes sense for those lights.

2. For lights that don't have approximately continuous spectrums (sodium vapor, some fluorescents, etc) dual illuminant profiles won't really help, because color temperature starts to be fairly useless in isolation - to get a good while balance, you're going to need tint as well. So if the image is important, create a single illuminant profile with that light source.

Finally, bear in mind that dual illuminant profiles are just an optimization - so far as I am aware, no other camera company/raw developer has found the improvement compelling enough to adopt the idea.

Sandy

« Last Edit: May 14, 2012, 12:01:42 PM by sandymc » Logged
digitaldog
Sr. Member
****
Online Online

Posts: 8604



WWW
« Reply #38 on: May 14, 2012, 12:20:18 PM »
ReplyReply

So in the situation where you shine a tungsten light on a white sheet of paper, what you're measuring is the color of the tungsten light, not that the sheet of paper is white.
If I understand what you’re saying, this would be akin to measuring the color of the light with say an EyeOne Pro using ambient head. IOW, you’re mesuring just the light, not the effect of the paper.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
sandymc
Full Member
***
Offline Offline

Posts: 248


« Reply #39 on: May 14, 2012, 12:50:40 PM »
ReplyReply

If I understand what you’re saying, this would be akin to measuring the color of the light with say an EyeOne Pro using ambient head. IOW, you’re mesuring just the light, not the effect of the paper.

Exactly so.

Sandy
Logged
Pages: « 1 [2] 3 4 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad