EM-5: Confused by Auto White-Balance results

Started Feb 12, 2013 | Discussions
Timur Born
Senior MemberPosts: 3,987
Like?
EM-5: Confused by Auto White-Balance results
Feb 12, 2013

Hey everyone!

I mostly do family snapshots, which means no controlled light and no white-cards around. Since I am shooting RAW I am not really depending on Auto White-Balance, but that doesn't mean I want to manually edit every shot.

And this is where the E-M5 (and other cameras) still leave me confused by their range of white-points being too broad and especially skin-tones being all over the place. Not to mention that the camera and each RAW software give different WB numbers to the same image file even when all are set to "as shot".

Here an examples of where I struggle to understand the differences. Both images were taken within minutes of each other, same target, same light (daylight through a window from the right image side), no indoor lights, no flash.

Reported "as shot" WB by LR4: 3600K

Reported "as shot" WB by LR4: 4150K

I've got several versions (shot within seconds) of both 3600K and the 4100K images in this series. Two possible explanations come to my mind:

1) Maybe some clouds may have hidden/revealed the sun within those 4 minutes time difference?

2) The 3600K images were shot at F1.8 and ISO 1600-2000, the 4100K images were shot at F6.3 and ISO 12800. This shouldn't matter for WB, but maybe it did.

Here another example, where I know the difference, but still struggle to accept that the outcome is so vastly different. Both shots were done within minutes, indoor, with no window/daylight, halogen lights (2900-3000K) coming straight from above, white/grey/black walls, white ceiling.

No flash, reported "as shot" WB by LR4: 2400K (shadow from my arm blocking the ceiling lights)

FL-600R pointed toward ceiling (-1 EV flash comp), reported "as shot" WB by LR4: 5150K

Again there is an ISO difference, ISO 2000 for the non-flash version, ISO 200 for the flash version. But again I wouldn't expect that to make much of a difference, at least not such a big one as seen here.

While I know that the second image is "mixed" light in that it mixes the flash with the halogen ceiling light my understanding of "Auto" WB is that it should get roughly the same results when presented with the same color targets. Obviously the skin and everything else looks vastly different here.

Since I had to push exposure by 2 stops in LR4 for the second shot I can only speculate that the flash was too weak while the camera expected it to fill in stronger versus the halogen light. Still I would expect "Auto" WB to measure the light, not just turn to 5000K once a flash is fired. Especially since there is a dedicated "Flash" WB setting, too.

Any hints and explanations welcome!

micksh6
Senior MemberPosts: 2,560Gear list
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to Timur Born, Feb 12, 2013

Timur Born wrote:

2) The 3600K images were shot at F1.8 and ISO 1600-2000, the 4100K images were shot at F6.3 and ISO 12800. This shouldn't matter for WB, but maybe it did.

I think it's mostly because lenses were different. Different lenses have different color signatures and AutoWB compensates for it.

Also flash will affect WB significantly depending on intensity and on how you bounce it (i.e. angle and flash position). Ceiling usually is not perfectly neutral, it will affect WB too.

Light reflected from walls, colorful clothes and other objects will also add to the mix. Even without flash, change camera angle and the mix of light reflected from various objects will change, affecting WB.

 micksh6's gear list:micksh6's gear list
Olympus PEN E-PL5 Olympus E-M1 Olympus M.Zuiko Digital ED 9-18mm 1:4.0-5.6 Panasonic Leica Summilux DG 25mm F1.4 Olympus M.Zuiko Digital ED 40-150mm 1:4-5.6 R +6 more
Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to micksh6, Feb 12, 2013

But isn't "Auto" WB supposed to compensate for changes of lens coloring and especially changes of light (temperature)? In the end the sensor and thus the "Auto" WB algorithms look through the lens, too. So when different lenses let through different "whites" then the algorithms should compensate for it, should it not?

And of course flash changes color temperature, but again "Auto" WB should make up for it. Of course it isn't perfect, but such a gross difference as shown in my second pair of shots is far from what I would expect from an "Auto" system (2400K vs. 5100K).

Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to Timur Born, Feb 12, 2013

Thinking about it, in the first shot 2400K is off the 2900-3000K that these halogen lights are rated at. And the image is too cool. On the other hand the second image is obviously too warm (too reddish skin), so the mixed halogen (3000K) + flash (5500K) shot doesn't look too good either (too close to flash WB).

Reply   Reply with quote   Complain
s_grins
Veteran MemberPosts: 8,896Gear list
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to Timur Born, Feb 12, 2013

Timur Born wrote:

Thinking about it, in the first shot 2400K is off the 2900-3000K that these halogen lights are rated at. And the image is too cool. On the other hand the second image is obviously too warm (too reddish skin), so the mixed halogen (3000K) + flash (5500K) shot doesn't look too good either (too close to flash WB).

If skin tones are so important to you

and

you have time to play with custom K balance,

why not to use gray card before you a take shot? (you do not have to place the card into bath tub).

Auto WB is an auto, so it is not perfect. I never use auto WB indoors, and outdoors I prefer to use "SUN" or "SHADOW" or "CLOUD" WB, but not auto.

Next is a PP, but this is separate issue

-- hide signature --

Looking for equilibrium...

 s_grins's gear list:s_grins's gear list
Panasonic Lumix DMC-GH2 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G 20mm F1.7 ASPH Panasonic Lumix G Vario 45-200mm F4-5.6 OIS Sigma 30mm F2.8 EX DN +2 more
Reply   Reply with quote   Complain
JosephScha
Veteran MemberPosts: 4,235Gear list
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to s_grins, Feb 12, 2013

I agree about setting up custom white balance.  But since he did shoot Raw these should be easy to correct, or at least to make consistent.

-- hide signature --

js

 JosephScha's gear list:JosephScha's gear list
Panasonic Lumix DMC-G10 Panasonic Lumix G Vario 45-200mm F4-5.6 OIS Panasonic Leica Summilux DG 25mm F1.4 +4 more
Reply   Reply with quote   Complain
assaft
Contributing MemberPosts: 906Gear list
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to Timur Born, Feb 12, 2013

Very noticeable difference between the shots. I don't recall seeing such cases with my E-PL2. I guess there isn't much you can do since you have no access to the algorithm. It would be interesting to see what a grey card sample would tell about these scenes. You can capture it from/with different angels/lenses/iso+aperture combinations to see if at least the light reflected from the card has changed and to what extent. It's a simulation of course, just in an attempt to rule out the effect of some parameters.

 assaft's gear list:assaft's gear list
Olympus PEN E-PL2 Olympus OM-D E-M10 Olympus M.Zuiko Digital ED 9-18mm 1:4.0-5.6 Panasonic Lumix G 20mm F1.7 ASPH Panasonic Lumix G 14mm F2.5 ASPH +5 more
Reply   Reply with quote   Complain
s_grins
Veteran MemberPosts: 8,896Gear list
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to JosephScha, Feb 12, 2013

JosephScha wrote:

I agree about setting up custom white balance. But since he did shoot Raw these should be easy to correct, or at least to make consistent.

-- hide signature --

js

RAW: you need gray card anyway, even so you are taking RAW. You can set WB before taking shot, or you can place gray card next to the subject and later on use it as a reverence, but you need it.

Without gray card adjusting colors during PP is a tedious job, even so it is RAW.

Saturation is another source of WB problems

-- hide signature --

Looking for equilibrium...

 s_grins's gear list:s_grins's gear list
Panasonic Lumix DMC-GH2 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G 20mm F1.7 ASPH Panasonic Lumix G Vario 45-200mm F4-5.6 OIS Sigma 30mm F2.8 EX DN +2 more
Reply   Reply with quote   Complain
assaft
Contributing MemberPosts: 906Gear list
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to assaft, Feb 12, 2013

besides that, I assume you keep the warm tone option off.

 assaft's gear list:assaft's gear list
Olympus PEN E-PL2 Olympus OM-D E-M10 Olympus M.Zuiko Digital ED 9-18mm 1:4.0-5.6 Panasonic Lumix G 20mm F1.7 ASPH Panasonic Lumix G 14mm F2.5 ASPH +5 more
Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to s_grins, Feb 12, 2013

s_grins wrote:

If skin tones are so important to you

and

you have time to play with custom K balance,

why not to use gray card before you a take shot? (you do not have to place the card into bath tub).

Because in most of these family snapshot situations I neither have time nor do I like to incur my family's wrath by running after them with gray cards.

In the two examples I posted here we have:

1) Baby and mother fully prepared and dressed (warm) to leave the house (cold).

2) Baby and mother being busy enough getting the baby clean without the baby drowning and moaning (she is still getting used to being bathed).

Auto WB is an auto, so it is not perfect. I never use auto WB indoors, and outdoors I prefer to use "SUN" or "SHADOW" or "CLOUD" WB, but not auto.

I will quote myself: "Of course it isn't perfect, but such a gross difference as shown in my second pair of shots is far from what I would expect from an "Auto" system."

Next is a PP, but this is separate issue

Seems like I am stuck with manual PP if both camera and in software (LR4) Auto WB hit so far off the mark.

Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to JosephScha, Feb 12, 2013

JosephScha wrote:

I agree about setting up custom white balance. But since he did shoot Raw these should be easy to correct, or at least to make consistent.

Yes, but it would be nice if I had not to work so hard on white-balance corrections even with shots that are done within mostly the same parameters (especially the same target/framing). I hoped that someone would see something in the images that explains the obvious differences, so that I may be able to detect the culprit beforehand next time.

Would other cameras than the E-M5 get similar results?

Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to assaft, Feb 12, 2013

assaft wrote:

Very noticeable difference between the shots. I don't recall seeing such cases with my E-PL2. I guess there isn't much you can do since you have no access to the algorithm.

One thing I am asking myself is how much other camera than the E-M5 would differ in these scenes or if there is something about the scenes/examples that leads to just the same expected outcome. I am struggling with flash and flash/mixed lighting anyway, but the first example was purely lit via daylight coming through the window.

It would be interesting to see what a grey card sample would tell about these scenes. You can capture it from/with different angels/lenses/iso+aperture combinations to see if at least the light reflected from the card has changed and to what extent. It's a simulation of course, just in an attempt to rule out the effect of some parameters.

I hoped that someone else would already have gathered that intel about the E-M5's (+lens) WB behavior. Especially since I already analyzed so many aspects of this camera and shared the information that I thought this information might be available from others for a change.

That being said, I already did some WB analysis with a gray and white target when I compared different RAW converters in another thread. This was the curious outcome:



Different software handles the same WB setting (camera) quite differently and up to now I don't even understand why their ranges/numbers differ so much (50000K? seriously?), even when their visible results don't necessarily.

Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to assaft, Feb 12, 2013

assaft wrote:

besides that, I assume you keep the warm tone option off.

Yes, I usually keep it off, albeit I am not sure yet how much I like it or not. Of course in RAW this only matters as far as the starting point is concerned.

Reply   Reply with quote   Complain
Paul De Bra
Forum ProPosts: 11,341Gear list
Like?
Auto White Balance is magic, and magic can go wrong...
In reply to Timur Born, Feb 12, 2013

All camera makers struggle with the auto white balance feature. I must say that in general I'm happier with the AWB on my E-M5 than I was on the Canon cameras I had before.

AWB tries to guess what the lighting was in a scene. It really has nothing to go on. An image contains objects of different colors, and the colors look different depending on the lighting of the scene. AWB tries to "neutralize" an overall color cast that could be caused by colored lighting such as tungsten lights. Essentially AWB deviates from a standard daylight when it discovers there is more red than expected (tungsten lights) or more blue than expected (cloudy weather, shadow on a sunny day, or light at dawn or dusk). When you take a picture containing almost all blue AWB will tone down the blue (by using a high color temperature) and it will not look right. And when you take a picture containing almost all red AWB will also tond that down (by using a low color temperature). AWB wants the "overall" color of a scene to be closer to a neutral grey.

There is only one situation in which AWB can do something that is at least partly sensible: when the flash is used. Assuming most of the light is coming from the flash the camera knows what color the light of the flash is. So even when you do not use the flash-WB preset the camera has a good idea about the lighting of the scene. In all other cases it's an educated guess, on the scene's overall color or the color of some larger parts of the scene. I am actually amazed that AWB on any camera does as well as it does.

-- hide signature --

Slowly learning to use the Olympus OM-D E-M5.
Public pictures at http://debra.zenfolio.com/.

 Paul De Bra's gear list:Paul De Bra's gear list
Fujifilm FinePix F200EXR Olympus OM-D E-M5 Olympus M.Zuiko Digital ED 9-18mm 1:4.0-5.6 Panasonic Lumix G 20mm F1.7 ASPH Olympus M.Zuiko Digital ED 75-300mm 1:4.8-6.7 +4 more
Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: Auto White Balance is magic, and magic can go wrong...
In reply to Paul De Bra, Feb 12, 2013

Thanks for the explanation, Paul. I know how AWB works, but it's still appreciated. <-winking smiley here, which currently doesn't work on the forum

My experience is practice is somewhat limited, though. I expected that AWB (on the E-M5) would come up with more consistent results when the scene, frame, target and colors are mostly the same. Especially since I got the impression that the E-M5 got some good reviews concerning its white-balance (didn't specifically check for that, though). This is why I am asking for other experiences and hints around here, to get a better "picture" (pun intended).

As can be seen by my comparison table one cannot even completely rely on the features of raw software when a gray target is used. Seems like lots of manual work is still in order...

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,131
Like?
Re: Auto White Balance is magic, and magic can go wrong ...
In reply to Timur Born, Feb 12, 2013

Timur Born wrote:

Thanks for the explanation, Paul. I know how AWB works, but it's still appreciated. <-winking smiley here, which currently doesn't work on the forum

My experience is practice is somewhat limited, though. I expected that AWB (on the E-M5) would come up with more consistent results when the scene, frame, target and colors are mostly the same. Especially since I got the impression that the E-M5 got some good reviews concerning its white-balance (didn't specifically check for that, though). This is why I am asking for other experiences and hints around here, to get a better "picture" (pun intended).

As can be seen by my comparison table one cannot even completely rely on the features of raw software when a gray target is used. Seems like lots of manual work is still in order...

.

Iliah Borg wrote:

Does the camera really invent a per-picture custom curve for each
of the RGB channels?

Curves are pre-programmed, not invented on the fly.

http://forums.dpreview.com/forums/post/29765442

From one of several posts found by Iliah Borg which are quoted/listed in this previous post:

http://forums.dpreview.com/forums/post/42346863

Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: Auto White Balance is magic, and magic can go wrong ...
In reply to Detail Man, Feb 12, 2013

Thanks for the links, DM. Interesting stuff! But does that mean in practice that it's illusionary to expect similar outcome when shooting similar scenes with different lenses or flash even the target/image colors are still the same?

Reply   Reply with quote   Complain
Detail Man
Forum ProPosts: 15,131
Like?
Re: Auto White Balance is magic, and magic can go wrong ...
In reply to Timur Born, Feb 12, 2013

Timur Born wrote:

Thanks for the links, DM. Interesting stuff! But does that mean in practice that it's illusionary to expect similar outcome when shooting similar scenes with different lenses or flash even the target/image colors are still the same?

It sounds like it may well be if it is indeed the case that cameras are but a "juke box" of pre-programmed "canned" WB settings that the system "lurches" between based on some rather crude measurments made using the RAW-level RGB channels.

Iliah Borg likely "knows his stuff" on that. I came out of those investigations with far less trust in Auto WB schemes than many users seem to have. Indoor lighting with tricky skin-tones is likely to be much harder than outdoor naturally lit environments.

Have found that even a rough Custom WB calibration to an indoor (even mixed artificial/natural) lighting environment can get one closer that Auto WB may. However, I am not talking about portraits with caucasian skin-tones, etc.

Have you tried (rough) Custom WB calibrating (as opposed to Auto WB) ? I wonder if it might (sometimes) serve you better with WB Temp/Tint "starting-points" ?

I do very little RAW processing with portraits and caucasian skin-tones. In what I do (nature/landascape stuff), I usually find that simply centering the Tint, manually adjusting the Temp to get close, and then some fine-tuning of the Tint (with perhaps some small amount of interactive adjustment together with Temp) works well (to my eyes, for what I photograph).

I don't find (with my GH2) that I ever use Tints greater than -2 to +2 when processing with DxO Optics Pro 7.x. Two (2) in either direction can be plenty (if not too much) in producing Green or Magenta tinting that is not desirable. I usually use no Tint at all (unless necessary).

As you know, the Adobe applications use higher numbers (allowing for more resolution) in their Tint adjustment controls. Not aware of how they correspond to DxO numbers (but they always seem to be many multiples larger in numerical value in LR 3.6). RAW Therapee (as you must know) seems to be nuetral at +1.000 (lower values Green, higher values Magenta, I think).

Reply   Reply with quote   Complain
Timur Born
Senior MemberPosts: 3,987
Like?
Re: Auto White Balance is magic, and magic can go wrong ...
In reply to Detail Man, Feb 12, 2013

CWB is worth a try, albeit I often am happy enough if I can hit the exposure button just in time to catch a moment. One reason why I shoot RAW is that it allows this inexperienced shooter more errors while shooting.

Just to mention it: On my screen the very first example (baby looking straight at the camera) shows the best looking skin tone. All others are off.

Reply   Reply with quote   Complain
assaft
Contributing MemberPosts: 906Gear list
Like?
Re: EM-5: Confused by Auto White-Balance results
In reply to Timur Born, Feb 12, 2013

yeah, i understand, and I remember this table from a previous post of yours.

I also experienced these differences in the interpretation of WB when I compared LR4 and OV2. I usually see high tint values when I open my orf files in LR. And usually the sky get a purplish color.

I sometimes use a 2-stop ND filter and I noticed that it adds some subtle yellowish tint to my pictures. For some reason the AWB can't compensate for that. But I guess that in such a case the shift is constant so averaging over a few samples of a grey card with and without the filter in order to learn what the shift is will give the a good approximation of the filter effect. A fixed WB can also be helpful for that. I don't know what you can do in your case...

Fuji is acclaimed for having great ooc jpegs, do you find the x10's AWB better then the EM-5's?

 assaft's gear list:assaft's gear list
Olympus PEN E-PL2 Olympus OM-D E-M10 Olympus M.Zuiko Digital ED 9-18mm 1:4.0-5.6 Panasonic Lumix G 20mm F1.7 ASPH Panasonic Lumix G 14mm F2.5 ASPH +5 more
Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads