Why White Balance?

Joined
Mar 4, 2005
Messages
145
Location
Portland, OR, USA
Was lying in bed this morning thinking about perception of color, light and white balance...

Everything in our physical reality has 2 colors; actual color and percieved color. The camera sees actual color and the eye sees percieved color. The color of anything is generally a combination of 2 things; the surface and the light hitting that surface. Each plays a role in the color of that object. Since the eye changes the color of what we see to something besides the actual color we need a way for the images our camera captures to also be changed to match as closely as possible the percieved colors in our photo. But why? If the photo is an exact replica of the scenes actual colors then won't our eyes play the same trick on us when we view it as it did when we looked at the scene and adjust accordingly making the photo look the same?

Obviously this doesn't happen, but why not?
 
Joined
Jan 29, 2005
Messages
2,761
Location
nowhere
because when looking on poto the surroundings are lighted with different light then the scene, and our eyes refuse to make the adaptation
 
Joined
Jan 28, 2005
Messages
301
Location
Cary, NC
realeyz said:
If the photo is an exact replica of the scenes actual colors
A photo is *not* an exact replica of the scene. It is a product of the camera's sensor and the resultant a/d conversion to a digital file, plus the subsequent a/d and d/a conversions that occur to show the image on the computer monitor. And the d/a conversion (complicated by ink colors) that occurs when the image is printed. If *each* element in the chain is *exactly* profiled to match the original scene, then your photo would be an exact replica of the scene. But that doesn't occur in reality, so camera manufacturers provide an approximation, called white balance, and we profile our monitors and printers. That has the chance of getting close, but we're still faced with the human eye and its limitations and the brain which translates what we see into what we *want* to see. The bottom line is adjust the white balance to something that approximates the original scene and that you find pleasing to the eye. That's my story and I'm sticking to it.
 
Joined
Feb 21, 2005
Messages
609
Location
Ottawa, Ontario
But Iliah, it then comes back to Todd's original question, why bother with WB?

If I'm sitting here in my darkened dungeon looking at an image on an uncalibrated screen and then I print the image and drive 30 miles to make a rare venture into the sunlight, according to your statement, I'm going to see two incredibly different images. Yet if I get the WB right (a rarity for me I'll admit) and get the rest of the color stuff right (and have my monitor and printer calibrated) then the image I see in both situations will be pretty much the same to my eyes/brain.

It leaves me thinking Joe's explanation was closer to the mark, but I'm willing to be convinced otherwise. :)

Neil
PS: The above assumes I read everything correctly.
 
Joined
Jan 29, 2005
Messages
2,761
Location
nowhere
A photo is *not* an exact replica of the scene.
What is scene? Is it how you see it? Oh how I see it? - But we see differently.
It is a product of the camera's sensor and the resultant a/d conversion to a digital file, plus the subsequent a/d and d/a conversions that occur to show the image on the computer monitor.
One converson for display IMHO - d/a.
And the d/a conversion (complicated by ink colors) that occurs when the image is printed.
RGB values to number of droplets and there positions is not d/a convertion IMHO.
If *each* element in the chain is *exactly* profiled to match the original scene, then your photo would be an exact replica of the scene.
I'm afraid profile is not the same as white balance.
But that doesn't occur in reality, so camera manufacturers provide an approximation, called white balance
As far as I understand white balance is only a sort of chromatic adaptation from actual colour temperature of the scene divided by native colour temperature of filter/sensor array to "common" white point; thus maintaining neutrality.
 
Joined
Jan 28, 2005
Messages
301
Location
Cary, NC
Iliah said:
A photo is *not* an exact replica of the scene.
What is scene? Is it how you see it? Oh how I see it? - But we see differently.
It is a product of the camera's sensor and the resultant a/d conversion to a digital file, plus the subsequent a/d and d/a conversions that occur to show the image on the computer monitor.
One conversion for display IMHO - d/a.
And the d/a conversion (complicated by ink colors) that occurs when the image is printed.
RGB values to number of droplets and there positions is not d/a conversion IMHO.
If *each* element in the chain is *exactly* profiled to match the original scene, then your photo would be an exact replica of the scene.
I'm afraid profile is not the same as white balance.
But that doesn't occur in reality, so camera manufacturers provide an approximation, called white balance
As far as I understand white balance is only a sort of chromatic adaptation from actual colour temperature of the scene divided by native colour temperature of filter/sensor array to "common" white point; thus maintaining neutrality.
Iliah,
One of my curses in life is being an electrical engineer. I understand the underlying physics of digital photography and these are the facts:

1. Photons, focused by lenses, impinge on the camera's sensor, either filtered by a Bayer filter or not.
2. Those photons produce an analog response in the CMOS (or other semiconductor technology) sensor.
3. The analog signals from the sensor are converted to digital streams by an a/d conversion.
4. The digital representation of the scene (as presented to the camera lens in the form of photons) is then manipulated by the camera's DSP using settings for WB, Hue, Tone, etc.
5. The resultant digital representation is written to a CF card.
6. The CF card is read by a computer and the digital file is sent to a monitor in either digital form (DVI) or analog form. If the latter, a d/a conversion takes place in the computer. In both cases the monitor, either CRT or LCD, converts (yes, another d/a conversion) the signals into an image that your eye sees on the screen. If the monitor has been profiled properly, the screen image will likely be a close representation of the scene that was photographed.
7. When you have manipulated the image until it appears on the screen the way you want it to appear, you send the final digital file to a printer. The printer converts (yes, d/a conversion) the digital file into an ink representation. This is probably the most crude conversion of the entire process since the purity of the inks, the number of inks, the size of the droplets, etc. all are factors in the d/a conversion that takes place.
8. If the printer has been profiled (i.e., printed image appears to the eye to be the same as the screen image), then the print is likely to be a close representation of the scene that was photographed.

I apologize for the dissertation but this is my domain of expertise and I rely on science for more answers than I should. I don't even attempt to answer questions like "if a tree falls in a forest and no one is around, does it make a sound?" But if a tree is photographed in the forest and the scene is taken through these steps properly, the resultant print will *look like* the original scene within the limits of the camera (dynamic range, sensor accuracy, number of pixels, settings including WB, etc.), the computer/monitor (profile accuracy, LCD or CRT accuracy, etc.), and the printer (profile accuracy, ink purity and consistency, droplet size, etc.). It's a *very* complicated process from end to end and it's easy to screw up along the way. But if you start with an accurate white balance setting in the camera, you stand a better chance of getting to the finish line successfully.
 
Joined
Jan 29, 2005
Messages
2,761
Location
nowhere
6. you mentioned two conversions in your original post, d/a and a/d. Please explain.

6. No, it will not be accurate representation of the scene unless the colour transform is right. And that is much more complicated then just white balance. As an extreme example, I can get white balance right in BGR, and all neutrals will be neutral. But the image will be totally weired :)

7. What exactly is going in inkjet printer? Do you mean here "media conversion"? Because a/d chips are not there.

8. Printer profile helps to render scene the way it looked in soft proof only if the viewing conditions are the same the profile was created for.
 
Joined
Jan 29, 2005
Messages
2,761
Location
nowhere
Ron,

Remember what Jean Armand Duplessis Duke de Richelieu used to say :) - "do not judge hastily"
 
P

peterdove

Guest
I have often been in my office working on an image, printed it out and it looks great, walked out into the sunlight and thought DAMN! You have to print something with the viewing conditions in mind.

Peter
 
Joined
Feb 27, 2005
Messages
294
Location
Burbank, CA USA
I won't Iliah. Once I have gotten through with this construction nightmare I'll have time to devote to analyzing the software and the workflow. I'll take time to do so.

By the way, Joe is a research and design engineer with a high level of competence in several rather esoteric and complex fields. His mind is a well-travelled and intricate labyrinth.

Ron
 
Joined
Jan 29, 2005
Messages
2,761
Location
nowhere
Ron, the question the thread started with was - "If the photo is an exact replica of the scenes actual colors then won't our eyes play the same trick on us when we view it as it did when we looked at the scene and adjust accordingly making the photo look the same?"

I made two points.

First, viewing conditions are different, and chromatic adaptation of the eyes goes to ambient light. If accurate scene rendition without chromatic adaptation would be viewed in the same light which was present in scene, no problem. But imagine an accurately rendered image of 9000 or 2800 scene without chromatic adaptation in software to 5000-6000K viewed at average daylight.

Second, camera does not record scenes colorimetrically, thus the resulting recording is not accurate (mostly photometrically for now).
 
Joined
Jan 28, 2005
Messages
301
Location
Cary, NC
Iliah said:
you mentioned two conversions in your original post, d/a and a/d. Please explain.
light conversion to digital stream - a/d
presentation of digital image on monitor - d/a (the eye sees analog, not digital)
presentation of digital image on paper using ink - d/a (once again, a picture is an analog representation, not digital)

6. No, it will not be accurate representation of the scene unless the colour transform is right. And that is much more complicated then just white balance. As an extreme example, I can get white balance right in BGR, and all neutrals will be neutral. But the image will be totally weired :)
Right on. It's more complex than just white balance.

7. What exactly is going in inkjet printer? Do you mean here "media conversion"? Because a/d chips are not there.
An inkjet printer takes a digital stream, over a USB or FireWire link typically, and converts it into micro-pulses of different color inks. By its very nature it is an a/d conversion. Unless your printer prints "1's" and "0's" it is converting a digital signal to an analog representation. Once again, the eye sees analog, not digital.

8. Printer profile helps to render scene the way it looked in soft proof only if the viewing conditions are the same the profile was created for.
Correct. That's why you should usually view your prints using an Ott light or in a light that is similar to the light of the original scene. Since the latter is not normally possible, I use an Ott light.
 
Joined
Jan 29, 2005
Messages
2,761
Location
nowhere
I was referring to this:
joecolson said:
plus the subsequent a/d and d/a conversions that occur to show the image on the computer monitor
joecolson said:
presentation of digital image on paper using ink - d/a (once again, a picture is an analog representation, not digital)
It is the concersion in our eye/brain more or less. Most inkjets run PWM and steppers for ink shots and positioning. Same as perforating tape, which is digital media :)
 
Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Nikon Cafe is a fan site and not associated with Nikon Corporation.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Copyright © 2005-2019 Amin Forums, LLC
Top Bottom