Pages: [1]   Go Down

Author Topic: 5D Mark II RGB Histogram accuracy  (Read 10144 times)

cp1

  • Newbie
  • *
  • Offline Offline
  • Posts: 5
    • http://www.distantlight.info
5D Mark II RGB Histogram accuracy
« on: March 31, 2009, 10:46:45 pm »

Following on from another topic concerning camera makers omission of a feature to display an accurate RAW histogram for image review, has anyone had any experience improving the accuracy of the RGB histogram shown on the 5D Mark II?

My old camera didn't have an RGB histogram and occasionally channels were getting blown out when setting the exposure purely on the luminance histogram.  So, when I got my 5D Mark II recently, I was optimistic that this would no longer be a problem.  Unfortunately it seems that the RGB histogram on the camera often shows channels being blown out when, upon inspection of the histogram for the raw image in Lightroom, this isn't actually the case.  So it appears that I've simply traded and over exposure problem for an under exposure one!  :-)

Has anyone had any experience improving the accuracy of the RGB histogram in relation to the captured raw data?  In particular, has anyone found that if any particular picture style gives more accurate results?  Or has anyone had any luck creating their own picture style to address this issue?

Cheers,

Christian.
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #1 on: March 31, 2009, 11:09:33 pm »

Quote from: cp1
Has anyone had any experience improving the accuracy of the RGB histogram in relation to the captured raw data?
There is not any "inaccuracy" there. The histograms are not supposed to reflect the raw data but the result of the in-camera conversion in JPEG, which takes place even if raw data gets recorded.

The histograms are based on the JPEG preview embedded in the raw file. If you want the JPEG based histograms resemble the raw histograms, you have to coax the camera into performing a "neutral" conversion, i.e. creating a JPEG preview with RGB channels resembling the raw channels. This can be done quite good, at the price that the preview becomes useless except for selection: most shots will be greenish, unsaturated and contrastlose. This is the price of perfect ETTR.

There have been many threads about this, some here too.- Here are some links to DPP threads, which link back to LL at some point.

http://forums.dpreview.com/forums/read.asp...essage=26905476
http://forums.dpreview.com/forums/read.asp...essage=26534648
http://forums.dpreview.com/forums/read.asp...essage=27554522

One aspect is not mentioned in those threads: that the color space should be sRGB.

Note, that the RGB like histograms do not help create perfectly ETTRed shots; they help judging if the exposure is perfect, i.e. the solution is in multiple shots, until you have an optimal one.
Logged
Gabor

button

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 427
    • http://
5D Mark II RGB Histogram accuracy
« Reply #2 on: April 01, 2009, 10:21:09 pm »

Hello Gabor,

Would you mind answering a few basic questions for me, regarding this thread:  http://forums.dpreview.com/forums/read.asp...essage=26534648


Once I have the shot of my monitor of Gluik's color chart in RAW analyze, I found the coordinates for 64/64/64.
   
 1) In the above thread, you state that one should "pass the image into photoshop" without any corrections.  How does one pass a RAW file into PS without ACR applying a white balance?  Surely, I'm missing something here.

 2) Also, I cannot correlate the RAW analyze coordinates in PS- there seems to be a huge mismatch.  How do you do that?

Thanks,
John
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #3 on: April 01, 2009, 10:39:52 pm »

John,

before you do any circumstantial work, try the simplest method (it is mentioned somewhere later): make a *totally blown* shot. Do not shoot the sun directly, but for example a white wall six stops higher exposed than metered. Try to use that image as WB template; this works with many cameras.

Some cameras act strangely; the above method does not work with all cameras (I think someone with al old 5D tried it and the camera did not accept the template, but I may be wrong with the camera); however, if it works, this is the best and easiest method.

Which camera do you have?
Logged
Gabor

Kirk Gittings

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1561
    • http://www.KirkGittings.com
5D Mark II RGB Histogram accuracy
« Reply #4 on: April 01, 2009, 11:15:14 pm »

I would love to have a more accurate histogram of the raw file. Having said that, photography for me (30+ years as a professional) has always required previsualization of a scene into a print, interpretation of a Polaroid etc. I find the histogram no different and am learning to look at it with the growing knowledge of how it differs predictably from the raw data.
Logged
Thanks,
Kirk Gittings

button

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 427
    • http://
5D Mark II RGB Histogram accuracy
« Reply #5 on: April 01, 2009, 11:58:49 pm »

Quote from: Panopeeper
John,

before you do any circumstantial work, try the simplest method (it is mentioned somewhere later): make a *totally blown* shot. Do not shoot the sun directly, but for example a white wall six stops higher exposed than metered. Try to use that image as WB template; this works with many cameras.

Some cameras act strangely; the above method does not work with all cameras (I think someone with al old 5D tried it and the camera did not accept the template, but I may be wrong with the camera); however, if it works, this is the best and easiest method.

Which camera do you have?


I have a 5d mark2.
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #6 on: April 02, 2009, 12:12:39 am »

Quote from: button
I have a 5d mark2.
In the meantime I checked it out; it was the 5D rejecting the template with the fully saturated shot. Give it a try, it takes only a minute. Do not use 1/3 stop ISO for this purpose (generally, there is no reason to use 1/3 stop ISOs). Pls post the result. If the camera accepts that shot as WB template (function "custom WB"), then switch the WB selection to "Custom" and make a shot. Rawnalyze displays the WB coefficients on the histogram panel; if you don't use Rawnalyze, upload one raw file and I take a look at it.
Logged
Gabor

button

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 427
    • http://
5D Mark II RGB Histogram accuracy
« Reply #7 on: April 02, 2009, 12:35:41 am »

Quote from: Panopeeper
In the meantime I checked it out; it was the 5D rejecting the template with the fully saturated shot. Give it a try, it takes only a minute. Do not use 1/3 stop ISO for this purpose (generally, there is no reason to use 1/3 stop ISOs). Pls post the result. If the camera accepts that shot as WB template (function "custom WB"), then switch the WB selection to "Custom" and make a shot. Rawnalyze displays the WB coefficients on the histogram panel; if you don't use Rawnalyze, upload one raw file and I take a look at it.

Hello Gabor,

I tried the "blown out" method, and the camera did in fact accept the template.  Unfortunately, the green channel is significantly overemphasized, leading to substantial underexposure, if the live histogram is utilized.  The preview image looks lime green, not the yellow-green that I've seen from those cameras whose white balance that has been properly adjusted for this procedure.  I'd upload the file, but I don't think 20 megabytes would fit (and I don't think the admins here would be too happy about it).  What would you suggest?

Thanks,
John
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #8 on: April 02, 2009, 12:43:42 am »

John,

use yousendit.com; if the file is under 25 MB, you don't need even to register (from 25MB to 100MB you need a free registration). You can use fictional email addresses as sender and recipient. After the uploading finished, you receive a page with a URL for the downloading, if you post it here, everyone can download it. If you don't want to publish the shot, send the URL to me in a message.

It is all right if the green appears much stronger than the other channels.

Have you made the other adjustments as well? Everything neutral? The "neutral" style is initially neutral. Note, that contrast and saturation have to be in the "middle" (setting 0), but sharpness 0 is at the left end of the scale.

ADDED: please upload the WB template file as well for verification. That should be much smaller.
« Last Edit: April 02, 2009, 12:44:57 am by Panopeeper »
Logged
Gabor

button

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 427
    • http://
5D Mark II RGB Histogram accuracy
« Reply #9 on: April 02, 2009, 01:21:06 am »

OK, here are the links:



http://www.yousendit.com/download/UmNKWWV0NmNGOFIzZUE9PQ       This is the overexposed shot.

[attachment=12666:WBCalibr...uillermo.jpg]  This is the calibration chart provided by Guillermo.

I am using the sRGB space, with picture style set to "neutral", with 0 sharpening, 0 contrast, 0 saturation, and 0 color tone.  The "blow out" shot was around 7 stops overexposed.


Thanks for the help, Gabor!  If this works, my live histogram will become all the more useful.

John
« Last Edit: April 02, 2009, 01:21:48 am by button »
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #10 on: April 02, 2009, 02:01:45 am »

John,

the blow-out shot is ok; however, something is strange. The template shot (the blown-out one) itself does not need to and normally will not appear as white, because that was shot with some other WB setting; that plays no role. However, this template has the WB coefficients R=1.03, G=1.0, B=0.95. There is nothing wrong with that, but how did you achieve this? Normally, i.e. if you shoot it with for example "daylight", you get some very different WB coefficients. Have you used Guillermo's method? Then you have already a relative good template, but you may be able to get a better one with the overexposed shot.

In other words, that WB setting, which was active when creating the blown out shot is suitable for ETTR (the 0.95 is not stellar, but ok).

I'm afraid you misunderstood some step; I rather explain it once more.

1. You create a shot, which will be used as WB template in the subsequent shots. It absolutely does not matter, with which WB setting you created the template shot (the totally overexposed one or the colorful one).

2. You declare that image as WB template through the function "Custom WB" (see page 66 of the manual). Select the function and select the image (the blown out one) and set it. Now the camera analyzes that image data, calculates and records the WB coefficients as "Custom WB".

There is an interesting note on page 67: if the exposure obtained in step 1 is way off, a correct white balance might not be obtained I don't know if this means that the camera will reject some templates, or that the result is not all right.

3. From now on, any time you select the WB "Custom" (this is not the function "Custom WB" but the big dial after you pressed the WB button), those WB coefficients will be used.

Make a shot of anything now, and upload that please.

If this works, you will get perfect WB coefficients and you don't need to do anything with Guillermo's method.

(I may go to bed in the meantime.)
Logged
Gabor

button

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 427
    • http://
5D Mark II RGB Histogram accuracy
« Reply #11 on: April 03, 2009, 06:27:14 pm »

Quote from: Panopeeper
John,

the blow-out shot is ok; however, something is strange. The template shot (the blown-out one) itself does not need to and normally will not appear as white, because that was shot with some other WB setting; that plays no role. However, this template has the WB coefficients R=1.03, G=1.0, B=0.95. There is nothing wrong with that, but how did you achieve this? Normally, i.e. if you shoot it with for example "daylight", you get some very different WB coefficients. Have you used Guillermo's method? Then you have already a relative good template, but you may be able to get a better one with the overexposed shot.

In other words, that WB setting, which was active when creating the blown out shot is suitable for ETTR (the 0.95 is not stellar, but ok).

I'm afraid you misunderstood some step; I rather explain it once more.

1. You create a shot, which will be used as WB template in the subsequent shots. It absolutely does not matter, with which WB setting you created the template shot (the totally overexposed one or the colorful one).

2. You declare that image as WB template through the function "Custom WB" (see page 66 of the manual). Select the function and select the image (the blown out one) and set it. Now the camera analyzes that image data, calculates and records the WB coefficients as "Custom WB".

There is an interesting note on page 67: if the exposure obtained in step 1 is way off, a correct white balance might not be obtained I don't know if this means that the camera will reject some templates, or that the result is not all right.

3. From now on, any time you select the WB "Custom" (this is not the function "Custom WB" but the big dial after you pressed the WB button), those WB coefficients will be used.

Make a shot of anything now, and upload that please.

If this works, you will get perfect WB coefficients and you don't need to do anything with Guillermo's method.

(I may go to bed in the meantime.)


Sorry for the delay, Gabor.  I've been playing around with a few settings.  Unfortunately, I don't have my camera with me at the moment.  When you are checking the RBG coefficients with with RAW analyze, what are you looking at, specifically?  When I pull up its histogram function, all I see are three separate RGB peaks with no numbers.

Thanks,
John
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #12 on: April 03, 2009, 08:07:54 pm »

Quote from: button
Sorry for the delay, Gabor
The only reason to be sorry if you don't succeed :-)

Quote
When you are checking the RBG coefficients with with RAW analyze, what are you looking at, specifically?  When I pull up its histogram function, all I see are three separate RGB peaks with no numbers.
See the highlighted numbers in the attached capture, from your totally blown shot.
Logged
Gabor

button

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 427
    • http://
5D Mark II RGB Histogram accuracy
« Reply #13 on: April 04, 2009, 10:18:53 am »

Quote from: Panopeeper
The only reason to be sorry if you don't succeed :-)


See the highlighted numbers in the attached capture, from your totally blown shot.

Ah... thanks.  So, ideally, all 4 of those numbers should = 1?  Why are there two green channel numbers?

Thanks again,
John
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #14 on: April 04, 2009, 11:22:28 am »

Quote from: button
So, ideally, all 4 of those numbers should = 1?
Right.

Quote
Why are there two green channel numbers?
The two green-filtered pixels in the color filter array are always alternating in line and column. Depending on the electronics, they might have slightly different characteristics, although I have not seen such differences yet, which would justify different WB values. Theoretically they could have different filters (perhaps they should). Thus some cameras' metadata contains independent, but in praxis always identical coefficients.
Logged
Gabor

jtrujillo

  • Newbie
  • *
  • Offline Offline
  • Posts: 37
    • http://jtrujillo.net
5D Mark II RGB Histogram accuracy
« Reply #15 on: April 05, 2009, 06:33:28 am »

Hi, did the overexposed shot with a 5DII and it's working amazingly precise! (When setting the custom wb to that overposed shot the camera warns that it may not work well with the shot but gives the option to accept it).

First I had some inconsistencies but I had the picture style to normal. When i set it to Neutral (all the values 0) the on-camera histogram matches 100% the histogram in C1.

For what is worth, I installed Rawnalyce (nice tool!) and it gives me Shot WB Coeffs RGGB 1.0313 1.0000 1.0000 0.9952. Current WB RGGB all 1.000 (The overexposed shot was on a white wall, 6 stos over the camera measurement, at sunlight.

Regards
Logged
juan htt

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
5D Mark II RGB Histogram accuracy
« Reply #16 on: April 05, 2009, 11:22:36 am »

For those willing to try the entire procedure  for UniWB (try first the quick method explained in the last section), I have just translated into English the original article here:

UNIWB. MAKE CAMERA DISPLAY RELIABLE.

BR
« Last Edit: April 05, 2009, 11:30:06 am by GLuijk »
Logged

Panopeeper

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1805
5D Mark II RGB Histogram accuracy
« Reply #17 on: April 05, 2009, 11:35:19 am »

Quote from: jtrujillo
When i set it to Neutral (all the values 0) the on-camera histogram matches 100% the histogram in C1
That's not a good sign. I don't know C1 at all, but I guess the displayed histograms reflect the result of the conversion. The in-camera histogram should look like the one displayed by Rawnalyze.

There are several actions you can not suppress by any setting. The color space conversion from the camera's color space usually in sRGB is unavoidable; bad, but that's life. The other factor is the non-linear mapping (gamma application). This one does not cause any mix between the channels, nor does it hide or cause clipping, but the histogram will not reflect the linear raw data.

The in-camera histogram should closely resemble the "Mapped" histogram of Rawnalyze (no WB applied, intensity adjustment 0, black point/white point as initially).
Logged
Gabor
Pages: [1]   Go Up