histograms, lenses and colour temp

[crisscross]

Following on from thread in DSLR by Greg/Perth43 Histograms and being unhappy with GUI's reply, I started looking for some landscapes where I had taken a number of exposures and also came across one where I had not changed speed or aperture, but had changed lens.

One where I had arguably/technically under-exposed 1/3EV came out very different (and worse) just adding .3EV in NEF/RAW exposure compensation.

More interesting was this one where I was probably trying to avoid a lens change from the 80-400 usually on when travelling, but realised I was going to have to. Shown with histograms in NX2.

No1: with the 80-400 at 86mm f11 1/320
No2: with 18-135 at 50mm f11 1/320


In order to get the colour temp correct at 5300± 'as shot' on no 2, on no 1 I have to change it from the 4700 registered to 5600.

No 3 is the normally edited version of no 2. On histogram only it could be judged underexposed .3 to .5EV, BUT upping it washes out the rest in a way that tightening the tone curve does not as it is not using the full sensor bandwidth or whatever you call it anyway.

Can someone tell me if it is normal for a lens change to have such a dramatic effect on colour temp? I don't think it always happens in fact I usually have to get the data up to see which lens I did use where it could have been either.

[dendrophile]

Hey Chris,

I am not an expert on this... but I guess you are asking how cameras auto-select the colour temp for a photo? Presumably cameras look through the pixels trying to find something that it can reasonably expect to be neutral-coloured, and works out the temperature from that. Quite how this is done I really have no idea. Your photos contain patches of sky and grass, so possibly the camera "knows" these are landscapes. It's not hard then to speculate that the camera could use the sky to get an idea of the colour temp. Given that the two photos contain different amounts of sky, it does not strike me as odd that it computes a different kelvin value. Hence I think it's not the lens change that has had this effect, rather the different framing that the lens has introduced.

Not sure if this helps any.

Regards,
Graham

[Colin Southern]

I've noticed this with Canon gear too - not sure what the root cause is. Classic example would be these two that I shot a few weeks ago - same SS / Aperture / ISO / Focal Length / WB - different results. On the face of it, only the lens was different.

Will need to do more testing when I have time.

[Dave Humphries]

I may be stating the obvious (and already known) here, but my understanding is this occurs due to differences in glass, and more particularly, the anti-flare coatings used during manufacture.

I suspect the outcome is mainly due to the different spectrums of vieling flare superimposed on the image, thus when measured in a lab with monochromatic light, the spectral responses of lens A vs lens B may well appear far more similar than real world results, like those above, show. (no broad-spectrum flare in the lab)

Just because they come from the same manufacturer doesn't exclude this because each lens generation (probably each batch) may have different coating formulae applied. That's all covered in the small print on the box that typically says the manufacturer reserves the right to continuous product improvement, meaning you can't rely on any two 'identical' products behaving the same!

Given all this, I'm not sure that even on auto WB, the camera could necessarily make them look the same.

I'm sure there's more to it than this, but that's my theory.

Cheers,

[crisscross]

thank you for your thoughts so far guys; looks like one of the great mysteries best solved by sliding sliders till one is happy! I suppose also at least in UK there are a lot of bits of thin cloud about that we don't notice in our ecstasy when sky seems blue and 20 secs between shots can allow one to float over the sun; Met Office satellite images often show things different from looking upwards.

Certainly colour temp seems 2nd only to exposure comp in adjustments worth looking at on RAW/NEF files

[Mike R]

It's normal for lenses to affect color temp. For example, see this post on Plant Neil. http://www.planetneil.com/tangents/2...-match-lenses/

I always thought that the color shift was due to the glass formulations (and additives), but I can't remember where I got that idea. It could also be due to coatings, as others have suggested.

[_GUI_]

Following on from thread in DSLR by Greg/Perth43 Histograms and being unhappy with GUI's reply.

I gave several replies in that thread, which one made you feel unhappy with it?

BR

[crisscross]

Hi BR, it was

Histograms

Partly because it is an extreme example, but more that my experience (using Nikon NX2) in PP is that correcting for exposure and/or colour temperature can also have a profound effect on saturation and contrast. This is on UK landscapes where there is often considerable haze and high humidity levels that need to be compensated for on an image in a way that the human eye does automatically/unconsciously.

It follows that there is also considerable benefit using Nikon software on Nikon camera shots as has a precise 'model' of the sensor-to-image preprocessing. Likewise DxO achieves a similar result by analysing the real-life output of all the cameras and lenses on its list.

[_GUI_]

I've noticed this with Canon gear too - not sure what the root cause is. Classic example would be these two that I shot a few weeks ago - same SS / Aperture / ISO / Focal Length / WB - different results. On the face of it, only the lens was different.

What is SS? anyway two lenses will never give a perfectly equal exposure in the RAW file, since the diapragm will not have exactly the same physical shape and size.

I have just done a quick test with the 24-70 f2.8L and the 70-200 f4L both shot at the same parameters (speed, f/8.0 aperture and ISO), and the shot from the 70-200 was a bit more exposed.
After a neutral development on both RAW files, exposure equalizing and contrast curve applied, the only differences I can see are those from the optical field (a slightly different FOV and geometrical distortion), and both histograms match perfectly:



BR

[_GUI_]

correcting for exposure and/or colour temperature can also have a profound effect on saturation and contrast.

I didn't say anything about white balance. White balance is independent from exposure; no matter which exposure you got in your RAW file, the appropiate WB will be the same for all exposures achieved.

Regarding colour changes for different exposures, it's mathematically impossible due to sensor linearity.
Just an example: if you shoot at 1/100 f/4.0 and ISO100, you get level X on the RAW file. If now you shoot at 1/50 f/4.0 and ISO100, you get level 2*X, i.e. one extra stop.
If you develop the first RAW applying +1EV extra exposure in the RAW developer, the software will just double all levels, i.e. multiply them by a factor of 2.0, taking us from X to 2*X level. The only difference will be in SNR which in the deep shadows is doubled with every extra exposure stop.

If you shoot with different exposures, and you get a different result in colours, contrast or saturation, the only explanation is that you didn't set both exposures equal before starting your postprocessing.

BR (this is not my name, this is just Best Regards )

[Colin Southern]

What is SS?

Sorry - by "SS" I was referring to "Shutter Speed"; I could probably have used a better abbreviation.

anyway two lenses will never give a perfectly equal exposure in the RAW file, since the diapragm will not have exactly the same physical shape and size.

Not sure if your meaning exposure as in "levels" or exposure as in "overall image". The difference I noticed in the image I posted was a change in colour, even with WB adjusted. The day was quite changeable (which I initially put it down to), but thinking back, it seemed like too big a change.

I'm not at all worried by it as it nulls out just fine with a gray card - just a little mystery that I'll investigate further some day.

[crisscross]

Hi Guillermo, and best regards

There seems to be a noticable difference between the 2 versions of your pic at my end, eg the red card more orangey at top and little brown jug richer generally

I brought in Colour temperature. No matter what the science says, I am not going to remove the feedback loop through my brain so as to discount what I have found in the past when doing PP on a similar case in the future. That possibly accounts for why I use Nikon camera, NX2 and process on mac, though I am sure Colin would agree with you that I should be using a Canon, processing in Photoshop and on a PC like every one else

Maybe sensors should be linear, lenses perfect etc etc, but the essential basis of DxO, which I think is aimed at serious professionals, is that analysis of actual output from various camera bodies and lenses reveals different and mildly aberrant effects - for which it will then auto-correct to save time. I certainly find DxO incredibly more effective on old 350D output than anything else.

[_GUI_]

Not sure if your meaning exposure as in "levels" or exposure as in "overall image". The difference I noticed in the image I posted was a change in colour, even with WB adjusted. The day was quite changeable (which I initially put it down to), but thinking back, it seemed like too big a change.

I'm not at all worried by it as it nulls out just fine with a gray card - just a little mystery that I'll investigate further some day.

I mean exposure in levels that affect the overall image but can be corrected at no cost. When you say 'even with WB adjusted' do you mean you used a different WB for each image? if you did, you did it wrong. The WB must be the same for both images in order to reveal any possible difference.

If the day was windy, just a few seconds in the moving clouds could make the difference. It was not a good example because differences could be both due to optics and the scene.

[Colin Southern]

I mean exposure in levels that affect the overall image but can be corrected at no cost. When you say 'even with WB adjusted' do you mean you used a different WB for each image? if you did, you did it wrong. The WB must be the same for both images in order to reveal any possible difference.

No - I used the same WB for those shots. What I was meaning by that other comment was any differences are easily nulled out when shooting regular (ie "non-test") scenes.

If the day was windy, just a few seconds in the moving clouds could make the difference. It was not a good example because differences could be both due to optics and the scene.

Could have been - no way of knowing without further testing. The shots came from a FOV test in a previous thread - I only used them in the thread you're refering to because it's all I had handy, and I thought it may help illustrate a point.

[_GUI_]

what I have found in the past when doing PP on a similar case

If you could just upload 2 RAW files in which you found differences in the final result because of a different exposure in the RAW files, I could try to reach some conclusions.

So far I have deeply analysed my 350D, and it's a _very_ linear device. Never had problems to make 2 images match in exposure just with a scaling factor applied to the RAW levels, and I am getting the same with the 5D I recently purchased.

Just an example: these 2 images came from a {-2,0,+2} bracketing and they are virtually equal afer exposure is adjusted. If you zoom 1:1 you'll simply see more noise in the 0EV version:




It is not a question of software, but hardware, luckily for us digital photographers.

BR