White Balance

[greepaul]

I took some photos on my Pentax K5 with the white balance set on automatic white balance (AWB) in continual shooting mode (several frames per second.) They were outdoor shots in sunshine (uncomplicated lighting conditions). In one series of three shots (just a fraction of a second apart), I noticed the colour balances were quite different and, looking at the raw files, I saw the colour temperatures were 5150, 5600 and 5650. Does anyone have any idea what could cause this?
Many thanks,
Paul

[NorthernFocus]

That's just the downside of letting the camera make decisions. I have three Nikon bodies and the same thing happens with them all.

[xpatUSA]

Was each scene identical in color and luminance content?

If not, AWB results can vary. And some cameras are "better" at AWB than others.

I've had very good results with Panasonic "G" series m4/3 and absolutely horrible results with Sigma SD14s.

[Shadowman]

Have a read of this tutorial, especially Notes about White Balance.

https://www.cambridgeincolour.com/tu...te-balance.htm

[rpcrowe]

If a moving object such as a person in a colored shirt moves in or out of a frame, the auto color can change. However, shooting in RAW, I have had no problems...

[xpatUSA]

If a moving object such as a person in a colored shirt moves in or out of a frame, the auto color can change. However, shooting in RAW, I have had no problems...

May I ask what difference shooting in raw makes?

I ask because I personally always shoot raw, but almost never use in-camera AWB, so it's not an area I'm familiar with.

[Mike Buckley]

May I ask what difference shooting in raw makes?

My thinking is that the "only" difference it makes is that it will be easier when using certain post-processing software to change the white balance. That's a big difference in my mind. When it comes to the camera selecting a different white balance when shooting in AWB mode, I can't imagine that the performance would be different based upon the file format being saved to the memory card.

[Mike Buckley]

I saw the colour temperatures were 5150, 5600 and 5650.

Let's pay attention to what really matters -- the look. Unless I'm comparing two images side by side, I very seriously doubt that I could tell the difference between 5150 and 5600 and I'm absolutely certain that I couldn't tell the difference between 5600 and 5650. So, aside from an interest in the technical details in and of themselves as part of a hobby, those differences wouldn't matter to me at all in the process of making images.

[xpatUSA]

Let's pay attention to what really matters -- the look. Unless I'm comparing two images side by side, I very seriously doubt that I could tell the difference between 5150 and 5600 and I'm absolutely certain that I couldn't tell the difference between 5600 and 5650. So, aside from an interest in the technical details in and of themselves as part of a hobby, those differences wouldn't matter to me at all in the process of making images.

I agree. A few hundred degrees didn't seem a lot to me either. So I compared three similar shots in RawTherapee - these images are all as-converted using the in-camera WB setting - here in Texas, that would be Sunlight of course







Scroll up and down and the first two are more or less the same but I moved for the third one.

Here's the temps and tints for each in order, after pushing the AWB button (didn't save though):

10,128K and 0.719
9,194K and 0.721
16,922K and 0.683

Food for thought, eh?

[FeatherMonkey]

My instinct is the sky in the final one. Whilst similar the final one is also different enough. Guessing middle grey is different enough in crude technical.

[xpatUSA]

My instinct is the sky in the final one. Whilst similar the final one is also different enough. Guessing middle grey is different enough in crude technical.

Yes, probably so. But what we should get from this test is that, by and large, AWB can be a hit or miss affair - whereas, intelligent observation of the scene before us with our own eyes should prompt the correct WB selection . . .

. . . or just set everything to Auto with 51-point evaluative metering, turn the big knob to 'P' and fire away

[IzzieK]

Buying a camera with the capabilities of you having the control on WB and ISO is already an expensive pocket exercise not to mention balancing your books. If you are going to choose Auto-everything, why not just get a point and shoot? Just my take on this conversion. In the meantime, I will suggest to Paul from Switzerland, whose post is not getting much traction, to upload the image/s that is causing him some problem/question...then we can go back to analyzing his post...what do you all think?

[Mike Buckley]

Buying a camera with the capabilities of you having the control on WB and ISO is already an expensive pocket exercise not to mention balancing your books. If you are going to choose Auto-everything, why not just get a point and shoot?

It will be interesting to see Paul's follow-up if he returns to the thread. However, he didn't mention that he used Auto-everything, only that he used Auto white balance. I bought two DSLRs more expensive than my previous cameras, one for me and one for my wife, partly so we could do exactly that and our cameras are rarely configured to any other white balance setting. That's because the more expensive, later-generation cameras have so much more accurate (effective) Auto white balance than the earlier generations.

So, I contend that when we spend more money on cameras with greater capabilities such as vastly improved Auto white balance, why not use those capabilities?

[Manfred M]

May I ask what difference shooting in raw makes?

I ask because I personally always shoot raw, but almost never use in-camera AWB, so it's not an area I'm familiar with.

Raw has no white balance (WB) / colour temperature value assigned to it, so anything goes. What we see in many editing tools is that the editor will select the white balanced recorded by the shot, depending on the camera settings as a starting position.

The user is free to adjust the WB to any value he or she would like as part of the raw conversion process.

[DanK]

Raw has no white balance (WB) / colour temperature value assigned to it, so anything goes. What we see in many editing tools is that the editor will select the white balanced recorded by the shot, depending on the camera settings as a starting position.

The user is free to adjust the WB to any value he or she would like as part of the raw conversion process.

Exactly. It's completely trivial to set WB to taste if you shoot raw. I always shoot raw, and I often leave the camera at AWB simply because the software (LR) reads it and gives me a starting point that is generally fairly close. For the same reason, I will sometimes choose another setting, e.g., a fixed WB when shooting at night or around sunset. However, it's just a convenience. I have never had the problem the OP had, but it would be trivial to fix: in processing, set the WB you want in one photo (or chose one that appears correct if there is one) and then sync that setting to the other photos. It's a matter of seconds in LR.

[NorthernFocus]

Raw has no white balance (WB) / colour temperature value assigned to it, so anything goes. What we see in many editing tools is that the editor will select the white balanced recorded by the shot, depending on the camera settings as a starting position.

The user is free to adjust the WB to any value he or she would like as part of the raw conversion process.

Which is why I don't understand all the fuss about WB to start with. OK, if one is shooting jpeg or aspires to in-camera processing (aka SOOC), then yes, it's an issue. But there's the solution isn't it. Shoot RAW and tweak it in post.

The OP asked WHY the AWB varies, not HOW to correct it. The WHY is because the camera is evaluating WB based on the color of light that it "sees". Subtle changes in the scene, as previously pointed out, will cause sometimes seemingly significant changes in the AWB setting. What we perceive to be the same scene isn't necessarily so to the camera. We have to remember that the camera is not evaluating ambient light but rather the light that is reflected by the scene that is falling on the sensor. Frankly because of this fact, I'm amazed at how well AWB works at all.

HOW to "fix" it if one desires control is to simply set WB manually in the camera rather than shooting AWB. It won't necessarily make things "right" but it does arguably simplify PP. At least that's one less thing to tweak if there is a series of shots that one desires to process identically.

[Mike Buckley]

We have to remember that the camera is not evaluating ambient light but rather the light that is reflected by the scene that is falling on the sensor.

Your overall point is correct but I think you misused the term, ambient. I think you meant to use the term, incident.

Frankly because of this fact, I'm amazed at how well AWB works at all.

Me too, especially considering that not that long ago in earlier-generation cameras it didn't work very well at all.

[NorthernFocus]

Your overall point is correct but I think you misused the term, ambient. I think you meant to use the term, incident...

Yes, just so.

[Manfred M]

Your overall point is correct but I think you misused the term, ambient. I think you meant to use the term, incident.

Me too, especially considering that not that long ago in earlier-generation cameras it didn't work very well at all.

I'm not quite sure that this is correct either, as the AWB is reading the reflected light of the scene. I know how on how the WB technology is likely to work, based on how wet darkroom colour analyzers worked. In some ways the issue is similar to why the reflective light meter in your camera will sometimes fail to give a correct light reading, whereas an incident light meter reading tends to be far more accurate.

Let me carry on the analogy a bit. A reflective light meter measures the amount of light falling on the metering cells in the camera and the camera manufacturer has assumed that any scene that is metered is the equivalent of "middle gray". For that reason, snow scenes end up being underexposed and end up looking too dark and low light scenes end up being too bright. Camera manufacturers know this and have built in exposure compensation functionality so that the photographer can override the camera system (based on his or her experience) and compensate for this inherent issue with reflective light metering.

AWB uses some of the same underlying logic; the assumption is that the overall image will have no predominant colour (i.e. if you mash all the colours in the image together, you end up with a neutral gray). This works fine in most images, but if there is a strong natural colour cast (say a tree of green leaves), the AWB will be fooled.

This is why photographers will shoot a colour chart of gray target so that they can perform a white balance in post-production. If you want to shoot jpegs, then you can do an in-camera custom white balance. You do need a large neutral target that is illuminated by the same light as your subject. Check your camera manual on how your camera does this. On my Nikons, this is integrated into the WB selection menu; I can't comment on the Pentax implementation.

[xpatUSA]

AWB uses some of the same underlying logic; the assumption is that the overall image will have no predominant colour (i.e. if you mash all the colours in the image together, you end up with a neutral gray). This works fine in most images, but if there is a strong natural colour cast (say a tree of green leaves), the AWB will be fooled.

I believe that Manfred is referring to the so-called "Grey World" method. There are also other methods, far more complicated. Generally, all these methods try to estimate the white point (not white level) of the scene illumination. By "white point" I mean the effective XYZ color of the illuminant which can be far, far different to "white" (D65) in sRGB color space. Once that white point is known, it is used to apply a transformation to the captured scene in order to correct it to the scene's estimated illumination. The transformation can be simple, as in 3 RGB multipliers (e.g. Canon) or more complex, as in a 3x3 matrix (Sigma).

Relevant to the discussion at hand is the fact that the Grey World, used in many cameras because of it's computational simplicity, is the worst possible method, with quite slim chances of success! So it is not surprising at all that the OP notes large differences between apparently similar scenes.

Here is a graph comparing various methods:




Note that the X-axis represents the number of surfaces in the scene. Chromaticity refers to a color as shown on a CIE horseshoe chart, most likely the Yuv model (not the xyY shown in all the simple color tutorials). So, if you shoot a two-surface scene, say sea and blue sky, AWB is going to get it quite wrong.

Here's some examples of the success of the various methods:



The images at left are raw with no AWB at all.

Here's the source of this information - very heavy reading indeed:

http://kronometric.org/phot/lighting...lancePaper.pdf

Basically, using AWB for in-camera JPEGs is asking for trouble, IMHO, and just clicking in a neutral area after the fact is not going to do a great job either.