Doug Kerr
Well-known member
This is the more-or-less annual lecture on fact and folklore in white balance color correction
************
A clever feature of human visual perception
The chromaticity of the light reflected from an illuminated surface depends on the concatenation of the spectrum of the illuminating light and the reflective spectrum of the surface. As an approximation we can say that the chromaticity of the light reflected from an illuminated surface depends on the interaction of the chromaticity of the illuminating light and the reflective chromaticity of the surface.
As a consequence, the chromaticity of the light reflected from a sheet of "white" paper illuminated by midday slight will be different than the chromaticity of the light reflected from the sheet when it is illuminated by late afternoon light.
Yet to our perception, the "color of the paper" seems the same. How can this be?
In fact, the human visual system "backs out" what it believes is the chromaticity of the ambient illumination on the subject (the paper), which it deduces from a survey of the chromaticity of the light from various familiar objects in the scene.
When the viewer does not see the subject in situ, but rather sees it in a photographic image, the human visual system again deduces the chromaticity of the incident light, but of course that is the incident light at the viewing location, not the incident light that fell upon the subject a year ago and a thousand miles away when the image was captured.
The human visual system "backs out" that chromaticity, and in general the result is that we perceive the white paper as "not white".
White balance color correction
To overcome this undesirable result, we use white balance color correction. We may ask the camera to apply this, or we may apply it in post-processing.
The principle is essentially to transform the chromaticities in the image to be the reflective chromaticism of the various scene elements. That will give the viewer the best chance to perceive the elements with their expected chromaticities. (We never say this, but this actually assumes that the viewer will be viewing the image under a certain assumed chromaticity of the surrounding environment.)
Essentially what we need to do this is knowledge of the chromaticity of the illumination upon the subject at the time the shot is taken.
Yes, cosines will be involved!
The chromatic photometer
A classical way to determine the chromatic of the illumination upon a subject is with a chromatic photometer (we may hear it called an incident colorimeter). Here we run into an issue that is parallel(!) to one in the matter of determining luminance for exposure planning purposes.
Suppose that the illumination on a subject comes from two sources with different chromaticity (perhaps daylight though a window and artificial light from a floor lamp).
The contributions of the chromaticities of those two sources to the net chromaticity of the "illumination" on the subject are weighted by the illuminance caused by each source on the subject surface. For the chromatic photometer to "follow along":
• Its photoreceptor must weight light arriving from different directions by the cosine of the angle of arrival - it must have a "cosine response".
• We must hold it so its photoreceptor is parallel to the subject surface.
For us, there are two disadvantages of the use of a chromatic photometer:
• They are expensive
• They typically report their findings of chromaticity in terms we cannot easily use to advise the camera, or the post-processing software, of what it needs to know.
Making the camera into a chromatic photometer
The white balance color correction measurement diffuser
Most modern digital cameras are prepared to cooperate in a scheme in which we make the camera temporarily into a chromatic photometer. This works especially handily when it is in fact the camera we want to know the chromaticity of the incident light, so it can do white balance color correction internally.
To do this, we mount on the front of the lens (as we would a filter) a white balance measurement diffuser. This takes the light incident on its surface and from it creates a (hopefully-uniform) luminous disk on its rear which the camera observes. The "acceptance" pattern of the diffuser should be of the cosine type. Thus (if the transmission of the diffuser is "chromatically neutral"), the chromaticity of the luminous disk presented to the camera will in fact be the net chromaticity of the illumination on the diffuser surface (regardless of how many separate sources, of differing chromaticity, might be involved).
We then hold the camera temporarily so that the face of the diffuser is parallel to the subject surface of interest and tell the camera to note the chromaticity of what is sees and use that as the premise for internal white balance color correction when we (a little later) take the actual shot.
We then remove the diffuser, return the camera to the shooting position, put it back into its normal "shooting" mode (telling it to use the most recently learned chromaticity as the premise for internal white balance color correction) and shoot away.
I'll return to the matter of measurement diffusers a little later.
The gray card
As an alternative to the use of a measurement diffuser, we can place a "gray card", whose reflectance is chromatically neutral, at the location of the subject surface of interest, and parallel to it. We then regard it with the camera (perhaps from the actual shooting position, although not necessarily) and ask it to note the chromaticity of what it sees. The rest of the scenario plays out as before.
Determination in post processing
Especially is we take the raw data output from the camera, it can be desirable to conduct the white balance color correction in the post processing software.
We could determine the chromaticity of the incident illumination with a chromatic photometer, and feed its findings into the raw processing software. But that is not straightforward.
A more practical approach is to include in a preliminary shot of the subject a "gray card" (held adjacent to and parallel to the surface of interest).
Then when we have loaded the raw files from this preliminary shot and the actual shots into our raw processing software, with the preliminary shot active, we point to the image of the gray card with the "eyedropper" or equivalent, telling the software, "The chromaticity of this spot is in fact the chromaticity of the incident light on out subject. Take note of it and use it as the premise for white balance color correction of the actual shots.
Another way to say that is: "This spot is 'white'. Behave accordingly."
An alternative way to use the camera as a chromatic photometer
The firm "ColorRight" offers, among other photographic accessories, a line of white balance measurement tools (the "ColorRight" tools). These mount in front of the lens (as we would mount a white balance measurement diffuser). They present a luminous disk to the camera as a result of the light falling on them.
With the tool mounted, and the camera at the position it will be for the actual shots, we aim the camera generally at the subject. As before, we ask the camera to note the chromaticity of what it sees, and to use that as the premise for internal white balance color correction on the actual shots to follow.
Wow! Very handy.
Now just how does this rig determine the chromaticity of the light incident on the subject surface?
Beats the hell out of me.
When this line of products first appeared (the name then was "Color Parrot"), I asked Drew Strickland, the promoter of the product, just how that happened. I commented that I was unable to construct any story based on optical or colorimetric theory that would explain how this camera-plus-tool would reliably become aware of the chromaticity of the incident light upon the subject surface by aiming toward the subject.
Drew, condescendingly, said there was no reason for nerds such as I to muddy the waters by inquiring into any "theoretical" considerations. Hundreds of people had bought the device, and that is the real story.
Now, how might it actually do what we need?
Roughly, the camera-plus-tool measures the chromaticity of the light incident on the tool. (Not rigorously, as its acceptance response is not "cosine".)
In many cases (especially outdoors, or perhaps in a "ballroom" setting), the illumination upon the photographer and the camera is very much the same as that on the subject. So, in such cases, if we sort-of measure the chromaticity of the illumination on the camera, that will be much like the chromaticity of the light on the subject surface, which is what we want the camera to learn.
Got that? Better lucky than good.
What about other cases?
Beats me.
Time here for breakfast.
Best regards,
Doug
************
A clever feature of human visual perception
The chromaticity of the light reflected from an illuminated surface depends on the concatenation of the spectrum of the illuminating light and the reflective spectrum of the surface. As an approximation we can say that the chromaticity of the light reflected from an illuminated surface depends on the interaction of the chromaticity of the illuminating light and the reflective chromaticity of the surface.
As a consequence, the chromaticity of the light reflected from a sheet of "white" paper illuminated by midday slight will be different than the chromaticity of the light reflected from the sheet when it is illuminated by late afternoon light.
Yet to our perception, the "color of the paper" seems the same. How can this be?
In fact, the human visual system "backs out" what it believes is the chromaticity of the ambient illumination on the subject (the paper), which it deduces from a survey of the chromaticity of the light from various familiar objects in the scene.
When the viewer does not see the subject in situ, but rather sees it in a photographic image, the human visual system again deduces the chromaticity of the incident light, but of course that is the incident light at the viewing location, not the incident light that fell upon the subject a year ago and a thousand miles away when the image was captured.
The human visual system "backs out" that chromaticity, and in general the result is that we perceive the white paper as "not white".
White balance color correction
To overcome this undesirable result, we use white balance color correction. We may ask the camera to apply this, or we may apply it in post-processing.
The principle is essentially to transform the chromaticities in the image to be the reflective chromaticism of the various scene elements. That will give the viewer the best chance to perceive the elements with their expected chromaticities. (We never say this, but this actually assumes that the viewer will be viewing the image under a certain assumed chromaticity of the surrounding environment.)
Essentially what we need to do this is knowledge of the chromaticity of the illumination upon the subject at the time the shot is taken.
Yes, cosines will be involved!
The chromatic photometer
A classical way to determine the chromatic of the illumination upon a subject is with a chromatic photometer (we may hear it called an incident colorimeter). Here we run into an issue that is parallel(!) to one in the matter of determining luminance for exposure planning purposes.
Suppose that the illumination on a subject comes from two sources with different chromaticity (perhaps daylight though a window and artificial light from a floor lamp).
The contributions of the chromaticities of those two sources to the net chromaticity of the "illumination" on the subject are weighted by the illuminance caused by each source on the subject surface. For the chromatic photometer to "follow along":
• Its photoreceptor must weight light arriving from different directions by the cosine of the angle of arrival - it must have a "cosine response".
• We must hold it so its photoreceptor is parallel to the subject surface.
What if the subject surface is not essentially flat? Well, then we may have a problem the meter can't cure.
For us, there are two disadvantages of the use of a chromatic photometer:
• They are expensive
• They typically report their findings of chromaticity in terms we cannot easily use to advise the camera, or the post-processing software, of what it needs to know.
Making the camera into a chromatic photometer
The white balance color correction measurement diffuser
Most modern digital cameras are prepared to cooperate in a scheme in which we make the camera temporarily into a chromatic photometer. This works especially handily when it is in fact the camera we want to know the chromaticity of the incident light, so it can do white balance color correction internally.
To do this, we mount on the front of the lens (as we would a filter) a white balance measurement diffuser. This takes the light incident on its surface and from it creates a (hopefully-uniform) luminous disk on its rear which the camera observes. The "acceptance" pattern of the diffuser should be of the cosine type. Thus (if the transmission of the diffuser is "chromatically neutral"), the chromaticity of the luminous disk presented to the camera will in fact be the net chromaticity of the illumination on the diffuser surface (regardless of how many separate sources, of differing chromaticity, might be involved).
We then hold the camera temporarily so that the face of the diffuser is parallel to the subject surface of interest and tell the camera to note the chromaticity of what is sees and use that as the premise for internal white balance color correction when we (a little later) take the actual shot.
We then remove the diffuser, return the camera to the shooting position, put it back into its normal "shooting" mode (telling it to use the most recently learned chromaticity as the premise for internal white balance color correction) and shoot away.
I'll return to the matter of measurement diffusers a little later.
The gray card
As an alternative to the use of a measurement diffuser, we can place a "gray card", whose reflectance is chromatically neutral, at the location of the subject surface of interest, and parallel to it. We then regard it with the camera (perhaps from the actual shooting position, although not necessarily) and ask it to note the chromaticity of what it sees. The rest of the scenario plays out as before.
Determination in post processing
Especially is we take the raw data output from the camera, it can be desirable to conduct the white balance color correction in the post processing software.
We could determine the chromaticity of the incident illumination with a chromatic photometer, and feed its findings into the raw processing software. But that is not straightforward.
A more practical approach is to include in a preliminary shot of the subject a "gray card" (held adjacent to and parallel to the surface of interest).
Then when we have loaded the raw files from this preliminary shot and the actual shots into our raw processing software, with the preliminary shot active, we point to the image of the gray card with the "eyedropper" or equivalent, telling the software, "The chromaticity of this spot is in fact the chromaticity of the incident light on out subject. Take note of it and use it as the premise for white balance color correction of the actual shots.
Another way to say that is: "This spot is 'white'. Behave accordingly."
An alternative way to use the camera as a chromatic photometer
The firm "ColorRight" offers, among other photographic accessories, a line of white balance measurement tools (the "ColorRight" tools). These mount in front of the lens (as we would mount a white balance measurement diffuser). They present a luminous disk to the camera as a result of the light falling on them.
With the tool mounted, and the camera at the position it will be for the actual shots, we aim the camera generally at the subject. As before, we ask the camera to note the chromaticity of what it sees, and to use that as the premise for internal white balance color correction on the actual shots to follow.
Wow! Very handy.
Now just how does this rig determine the chromaticity of the light incident on the subject surface?
Beats the hell out of me.
When this line of products first appeared (the name then was "Color Parrot"), I asked Drew Strickland, the promoter of the product, just how that happened. I commented that I was unable to construct any story based on optical or colorimetric theory that would explain how this camera-plus-tool would reliably become aware of the chromaticity of the incident light upon the subject surface by aiming toward the subject.
Drew, condescendingly, said there was no reason for nerds such as I to muddy the waters by inquiring into any "theoretical" considerations. Hundreds of people had bought the device, and that is the real story.
Now, how might it actually do what we need?
Roughly, the camera-plus-tool measures the chromaticity of the light incident on the tool. (Not rigorously, as its acceptance response is not "cosine".)
In many cases (especially outdoors, or perhaps in a "ballroom" setting), the illumination upon the photographer and the camera is very much the same as that on the subject. So, in such cases, if we sort-of measure the chromaticity of the illumination on the camera, that will be much like the chromaticity of the light on the subject surface, which is what we want the camera to learn.
Got that? Better lucky than good.
What about other cases?
Beats me.
Time here for breakfast.
Best regards,
Doug