Ad
Ad
Ad
Pages: [1] 2 3 »   Bottom of Page
Print
Author Topic: "Settings for an Accurate Histogram": WB?  (Read 15257 times)
joedecker
Full Member
***
Offline Offline

Posts: 142


WWW
« on: September 22, 2009, 01:08:40 PM »
ReplyReply

Nice article.  One question, though....

For per-channel histograms, on cameras that create a histogram even for RAW images based on the associated JPG, doesn't the white balance setting affect the relative brightnesses of the channels and therefore affect the accuracy of the histogram as well?

Logged

Joe Decker
Rock Slide Photography
http://www.rockslidephoto.com/
Andrew Fee
Jr. Member
**
Offline Offline

Posts: 87


WWW
« Reply #1 on: September 22, 2009, 01:19:46 PM »
ReplyReply

With Canon (at least on my 1000D) if you put the lens cap on, take a shot using the smallest aperture, lowest ISO and highest shutter speed and then use it to set the white balance, it sets the RGB multipliers to exactly 1, so you shouldn't have that problem.

It looks very ugly on the back of the camera though, and means JPEGs are basically unusable. (but you're not doing this to shoot JPEG anyway)

I really wish you could shoot with settings that gave you a rough approximtion of what your ‘developed’ RAW file will look like on the back of the camera, and for JPEGs, but that the histogram would be based on the RAW data rather than the JPEG preview.
Logged
NikoJorj
Sr. Member
****
Offline Offline

Posts: 1063


WWW
« Reply #2 on: September 22, 2009, 02:36:53 PM »
ReplyReply

Quote from: Andrew Fee
With Canon (at least on my 1000D) if you put the lens cap on, take a shot using the smallest aperture, lowest ISO and highest shutter speed and then use it to set the white balance, it sets the RGB multipliers to exactly 1, so you shouldn't have that problem.

It looks very ugly on the back of the camera though, and means JPEGs are basically unusable. (but you're not doing this to shoot JPEG anyway)
Yes, reading the introductory sentence on the what's new page, I was just awaiting an UniWB tutorial  . The tips gaved here are more basic, but maybe more usable too.  

To second Michael's conclusion note, it would be great that photographers don't need such tortuous techniques to simply assess what's in the raw data they captured - and going on on this, why no expose-to-the-right matrix metering when in raw mode?
Logged

Nicolas from Grenoble
A small gallery
knweiss
Jr. Member
**
Offline Offline

Posts: 71


« Reply #3 on: September 22, 2009, 03:04:35 PM »
ReplyReply

This contrast hint is interesting. I've now created a user defined setting on my Canon 5D2: Picture Style=Neutral, Sharpness=0, Contrast=-4, Saturation=0, Color tone=0.

The Picture Style seems to influence the in-camera histogram, too. I've selected "Neutral" but and I am not sure about the practical difference between "Faithful" and "Neutral". Which setting do you use if you shoot RAW-only?
« Last Edit: September 22, 2009, 03:05:40 PM by knweiss » Logged
Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1709


« Reply #4 on: September 22, 2009, 04:22:29 PM »
ReplyReply

Not sure I agree 100% with the article. Perhaps for applications where you will be putting significant effort into processing each picture it would make sense but when you are going to be applying a set minimum of contrast to your image as default then suddenly your histogram is not so accurate anymore because at zero contrast it is not true to the possibilities within the final image. Give you an example. I'm a wedding shooter. I need to process hundreds of photos with very high contrast/DR. Now if I set at zero contrast and shoot ETTR then when I apply the contrast needed for print I'm going to blow the dress. I of course do have that information and with the Local Adjustment Tool have huge amounts of 'recovery' room however the histogram is not accurate relative to the contrast curve of the final image which I work with when trying to process hundreds of images fast.

I set my camera to 1 click under 'neutral' contrast and I find that it gives me a histogram which accurately reflects the extra RAW highlight information compared to the jpg but doesn't give me an impression of more highlight headroom than I would have WITHOUT having to use dodge and burn tools in the RAW converter to get a finished result without having to enter PS.

I know that my specific circumstances may differ from the majority on this site who concentrate on 'each image will have it's own unique processing start to finish' type of photography but for people like me, the ability to see what your image will look like with the minimum amount of faffing around in post processing, both in the preview and histogram, is very valuable. I would never ever shoot jpg but neither do I shoot weddings with the kind of processing that my landscapes would require either. I have a default set of values in ACR, a custom profile or two, I change the WB and tweak the brightness (and exposure if highlights need recovering) and that is it apart from dodge and burn where absolutely necessary. Batch to jpgs and send to the lab for printing as proofs. ETTR is pretty much accurate exposure for wedding work anyway, make the dress white but holding detail and the face should be pretty much spot on.

Think what I'm trying to say is that if you need to process and get out a lot of prints, without being false to the idea of ETTR, it makes sense to try and get the exposure as spot on as possible in camera to reduce the amount of post work. You can't do that if your histogram thinks that you only want zero contrast.
« Last Edit: September 22, 2009, 05:16:49 PM by pom » Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #5 on: September 22, 2009, 05:42:11 PM »
ReplyReply

Quote from: joedecker
Nice article.  One question, though....

For per-channel histograms, on cameras that create a histogram even for RAW images based on the associated JPG, doesn't the white balance setting affect the relative brightnesses of the channels and therefore affect the accuracy of the histogram as well?

That is true, and many photographers load a UniWB into the camera so that the white balance multipliers are equal to unity, giving a better evaluation of the actual status of the raw data in the RGB channels. The black and white histogram is usually a luminance histogram and is most sensitive to green and least sensitive to blue; such a histogram may not demonstrate clipping in the red or blue channel. A composite RGB histogram (such as is default in Photoshop) would represent RGB equally and would be preferable in detecting blown red or blue channels. The Cambridge in Color site gives a good discussion of this topic. If Adobe RGB is the widest color space that the camera offers, then the histogram may show saturation clipping when none is present. It would help if the cameras offered a wide gamut space such as ProPhotoRGB.

Using a low contrast setting on the camera is often recommended and is helpful if the camera histogram gives a conservative estimate of clipping (as most do), and indicates clipping when none is present, allowing a safety margin for the highlights. Advocates of ETTR may not want this headroom. Increasing the contrast setting on the camera applies an S curve to the tonal data, lifting the three quarter tones and lowering the quarter tones. It does not affect the extreme highlights or extreme shadows. At the two extremes of the contrast curve, input equals output. If the camera histogram is conservative, then lowering the contrast will lower the three quarter tones and less clipping will be apparent in the histogtram. If the histogram is dead on, the contrast setting will have no effect. A better way of improving the appearance of the histogram would be to load a custom curve in the camera so that clipping in the raw channels takes place when the histogram is fully to the right. The contrast of the preview image would then be normal.

One should compare the raw histogram to the camera histogram to determine the amount of headroom that the camera gives, and make a corrective custom curve if indicated.
Logged
dreed
Sr. Member
****
Offline Offline

Posts: 1170


« Reply #6 on: September 22, 2009, 07:25:25 PM »
ReplyReply

It would be nice if the rendering of the preview image, displayed on the back of the camera, could be "exposure corrected" when shooting ETTR. When shooting, I've had people ask "was it a good pic?" and my reply is (almost universally) "I'll know when I load it onto a computer." This leaves them a bit puzzled because almost everyone expects the picture on the back to be what you've just taken...

The other by-product of ETTR is that I no longer pay any attention to review comments about back of the camera LCD screen colour accuracy (the image is not exposed correctly, so the rendered colours are going to be meaningless.) The only detail that's important, now, is whether or not I can get 100% zoom to check if it is properly focused.

A software option in the firmware to "correct the exposure" could be interesting...

Logged
Daniel Browning
Full Member
***
Offline Offline

Posts: 142


« Reply #7 on: September 23, 2009, 01:14:17 AM »
ReplyReply

Other settings for a more accurate histogram:

  • White balance and color space, as others mentioned, that's a big one.
  • Auto Lighting Optimizer (or whatever your cameras calls the context-sensitive local brightness adjustment)
  • Disable Highlight Tone Priority (or whatever your camera calls the setting that reduces ISO and applies nonlinear EC)
  • Tone curve: This is generally the same as contrast.
  • Gamma curve: if your camera allows loading a different gamma curve. (Linear is not useful in the general case IMHO, but would be nice to have a one-button push for comparing with normal gamma.)

I'm sure there are more I'm forgetting. All these are dumb workarounds for what all cameras should have already: a raw histogram. RED ONE is an example of the kind of additional tools that would be highly useful. One is a false color mode that shows every raw intensity as a different color, like Predator's infrared vision. That way you can see exactly what part of the image maps to what part of the histogram.
Logged

--Daniel
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1273



WWW
« Reply #8 on: September 23, 2009, 03:36:05 PM »
ReplyReply

This forum has a thread that is IMO more interesting and accurate to have the best possible information about the RAW file than this recently appeared article. It is here: Camera's histogram reliable to the RAW data, Specific WB for realistic camera JPEG. The detailed article here: UNIWB. MAKE CAMERA DISPLAY RELIABLE

I disagree with michael that the correct WB must be set to have the best knowledge about the RAW, because when applying this WB the camera performs overexposure over the RAW data to generate the JPEG that provides the LCD histogram. Because of this, the camera histogram is pesimistic telling us certain areas are blown that actually remain intact in the RAW file.

So the best way to have a more reliable histogram in the camera is not set the correct WB but cancel the WB. This will give us information about what we can really expect to find when developing the RAW file. The so called by Illiah Borg 'UniWB' is not perfect, but much better than using the correct WB or fiddling with the saturation, contrast or colour profile features of the camera (that anyway can be tuned at the same time as UniWB).

Regards.
« Last Edit: September 23, 2009, 03:38:52 PM by GLuijk » Logged

Panorama
Jr. Member
**
Offline Offline

Posts: 63


« Reply #9 on: September 24, 2009, 08:43:22 AM »
ReplyReply

An old trick and I've been doing this for years. I shoot Canon bodies and set the contrast, saturation, and color tone to -4 and sharpness to 0 (the lowest setting). This not only affects the histogram but in the event that you shoot jpegs you get the most control over your images; you can mod them in post as opposed to having the camera do something you may not want.

Same applies to Nikon; they're heavy handed on the saturation, so I'd definitely be turning the sat down at a minimum if I shot Nikon. Why give control to the camera?
« Last Edit: September 24, 2009, 08:45:48 AM by Panorama » Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #10 on: September 24, 2009, 10:09:52 AM »
ReplyReply

The best setting needs to be figured out for every camera model (and, ideally, for every kind of illumination, but that's not realistic). One should remember, that the neutral contrast and saturation settings are not the lowest values but zero. I find it strange, that some are using the minimum contrast and saturation setting.

Regarding curves: I know of no camera, which allows for loading a gamma curve. The curve one can load is a substitute for the standard contrast curve, or S-curve; a "linear curve" eliminates that transformation, i.e. it moves closer to the raw histograms.

However, the loadable curve can be used to compensate for the gamma curve. I created a custom curve for my 40D, resembling the inverse gamma curve. Note, that the curve never affects clipping, but it does affect the distribution of the intensities (the appearance of the histograms). The closer the displayed histogram resembles the raw histogram, the better one can judge the "exposure lattitude between the currect exposure and clipping.

Another important issue is the color space. This has to be tested with every camera model. I found, that my Canon 40D has to go with sRGB, otherwise the displayed histograms show lower exposure than the raw data itself. One could say it is logical, for AdobeRGB "compresses" the pixel values more, but it is not so simple. I suggest trying it out.
Logged

Gabor
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #11 on: September 24, 2009, 10:12:18 AM »
ReplyReply

Quote from: Panorama
An old trick and I've been doing this for years. I shoot Canon bodies and set the contrast, saturation, and color tone to -4 and sharpness to 0 (the lowest setting). This not only affects the histogram but in the event that you shoot jpegs you get the most control over your images; you can mod them in post as opposed to having the camera do something you may not want.

Same applies to Nikon; they're heavy handed on the saturation, so I'd definitely be turning the sat down at a minimum if I shot Nikon. Why give control to the camera?
I fail to see the logic in this reasoning. Negative contrast and saturation setting is just as a change of the image data as a positive value.
Logged

Gabor
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #12 on: September 24, 2009, 10:17:14 AM »
ReplyReply

I created a demonstration of ETTRing with my Canon 40D. I made two shots of the very same scenery (shrubs in my yard), in manual mode, same aperture, shutter and ISO, only a few seconds apart. The first one was a picture style changed to high contrast and high saturation, white balance "daylight"; the second one was made with neutral settings and a neutral custom WB (some call it UniWB), and with the inverse gamma curve.
 
I photographed the in-camera displays (sometimes it's good to have the old 20D in reserve):
 




The first one shows the sky as mostly clipped (see the black at the top left part?). The second one does not show any clipping by flashing, and the green channel's histogram shows a small clipping.

Now, the fact: clipping in all three channels, but very small (the y axis is logarythmic), only between the leaves; there flashing clipping indication on the sky is totally incorrect.



The following capture shows most of the clippings; magenta indicates green clipping, the red shows green and blue clipping and black shows, that all three channels clipped:



This shows, that the special setup reflects the real (raw) exposure very closely, while the other setup is useless for ETTR, because it suggests reducing the exposure.
Logged

Gabor
bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #13 on: September 24, 2009, 12:47:20 PM »
ReplyReply

Quote from: Panopeeper
I created a demonstration of ETTRing with my Canon 40D.
This shows, that the special setup reflects the real (raw) exposure very closely, while the other setup is useless for ETTR, because it suggests reducing the exposure.
An excellent demonstration, Gabor. I have done a few studies with histograms on my Nikon D3. For scenes without saturated colors and without saturation clipping imposed by white balance, the camera histogram wtih the standard picture control does a good job. Here is a Stouffer wedge with the green channel slightly below clipping:



And the camera histogram is also just short of clipping. Accurate enough for most work.



With a saturated yellow flower, the situation changes. The raw histogram shows the red channel at clipping and the green channel just short of clipping:



The camera histogram with the camera set to aRGB demonstrate strong clipping in green and red, but the luminance histogram appears OK. The green clipping occurs because the green does not fit into aRGB. The extreme red clipping is due both to the white balance multiplier and color space limitation. UniWB would have been helpful, but I was not using it at that time.



Reducing exposure will eliminate the green clipping when the green comes into the gamut of aRGB, but there will still be massive red clipping because of the red multiplier. This image shows decreasing exposure from top to bottom:



The screen shot below shows a decent rendering into ProPhotoRGB. Negative exposure is needed, since the original exposure was fully to the right. In this case, UniWB would have been helpful, but without a wider space than used for the aRGB preview, there would have been underexposure of the green.

« Last Edit: September 24, 2009, 12:57:37 PM by bjanes » Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 8093



WWW
« Reply #14 on: September 24, 2009, 01:13:59 PM »
ReplyReply

Quote from: Panopeeper
This shows, that the special setup reflects the real (raw) exposure very closely, while the other setup is useless for ETTR, because it suggests reducing the exposure.

Real or just closer? I’ve tried all kinds of settings on my Canon, some do produce a closer approximation of the Raw data but its pretty far off (calling it approximation is being kind).
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1709


« Reply #15 on: September 24, 2009, 01:20:18 PM »
ReplyReply

Quote
The screen shot below shows a decent rendering into ProPhotoRGB. Negative exposure is needed, since the original exposure was fully to the right. In this case, UniWB would have been helpful, but without a wider space than used for the aRGB preview, there would have been underexposure of the green.


Question: You aren't printing in ProPhotoRGB (can any printer do all of ProPhoto?) so I assume that when you convert to a smaller space some colours will start clipping again. As such would it not make more sense to try and equalise the histogram towards nearer towards the output colour space to reduce problems?

Or when you normalise the exposure from the original ETTR will you be anyway pulling the clipped colours back 'into the fold'?
« Last Edit: September 24, 2009, 01:20:38 PM by pom » Logged

Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #16 on: September 24, 2009, 01:52:20 PM »
ReplyReply

Quote from: digitaldog
Real or just closer?
There is no "real", for there are transformations involved, which we can not universally counter. Already the demosaicing causes shifting between the raw channels, but the biggest item (if the WB issue is solved) is the color space transformation. As this results in complex shiftings between the raw and the RGB channels, it can not be balanced (countered) by the minimum of adjustments available in-camera.

I can imagine, that different, specific not neutral WB templates, depending on the scenery (representing the prevailing source colors) and on the illumination would help in creating histograms more closely resembling the raw data, but the involved labour is probably enormous - and that would have to be repeated with every camera model, because that depends on the spectral characteristics of the sensor.
Logged

Gabor
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #17 on: September 24, 2009, 02:00:35 PM »
ReplyReply

Quote from: pom
would it not make more sense to try and equalise the histogram towards nearer towards the output colour space to reduce problems?
The need to reduce the "exposure" in raw processing excludes automatic raw conversion, but I don't see any other "problem".  Though I do realize, that this aspect can be serious for those, who are creating huge number of raw files and converting them in batch.

On the other hand, the higher real exposure reduces the noise, compared to a non-ETTR exposure, supposed that the DR of the scenery is large enough to justify ETTR in the first place.
Logged

Gabor
bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #18 on: September 24, 2009, 02:28:34 PM »
ReplyReply

Quote from: pom
Question: You aren't printing in ProPhotoRGB (can any printer do all of ProPhoto?) so I assume that when you convert to a smaller space some colours will start clipping again. As such would it not make more sense to try and equalise the histogram towards nearer towards the output colour space to reduce problems?

Or when you normalise the exposure from the original ETTR will you be anyway pulling the clipped colours back 'into the fold'?

Mapping the colors that are out of gamut for the printer or display device from ProPhotoRGB is another matter. Reducing the luminance will bring some colors back into gamut with a narrower output space, since the gamut for saturated colors is greatest at mid-luminances in an RGB color space, but that may not the best approach. The aim of the raw capture is to capture as much of the scene luminance and color as possible. You may not be able to print saturated colors at high luminance and a reflection print can not show the DR of the capture, but at least you can map those values to best advantage. Besides, output devices are improving all the time.
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #19 on: September 24, 2009, 02:30:54 PM »
ReplyReply

Quote from: digitaldog
Real or just closer? I’ve tried all kinds of settings on my Canon, some do produce a closer approximation of the Raw data but its pretty far off (calling it approximation is being kind).

Do Canon cameras permit rendering into a wider space than aRGB by the camera?
Logged
Pages: [1] 2 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad