Follow up from Nikon D2X training - exposure and the histogram

Discussion in 'General Technical Discussion' started by Luckynp, Sep 24, 2005.

  1. This is probably old information for most of you but it's for me it's the most important thing I got out of the Nikon training day for the D2X.

    I hadn't realised before how inportant it is to nail exposure in the camera. My method generally has been to avoid blowing hightlights and (secondarily) to try to get the histogram as far over to the right as possible. But often, I've settled for a histogram mostly over to the left and thought "I'll fix it in PS later". A phrase the trainer said was amongst those he hated most.

    The revelation (for me) was how much of the information resides in the right hand side of the histogram.

    Forgive me if I don't use technically correct terms and express myself badly - I'm not a technical-type person.

    As I understand it, it goes like this. If you mentally divide the histogram area into 6 zones (roughly corresponding to the 6 stops the camera can handle), then starting from the right, each zone contains twice as much information as the one to its left. This means that one half of all the available information is recorded in the 1 sixth area at the extreme right of the histogram. The distribution is as follows, starting from the left (darkest area) when shooting 12/16 bit.

    1 64 tones
    2 128 tones
    3 256 tones
    4 512 tones
    5 1024 tones
    6 2048 tones

    Adding up to the total 4096.

    So, if you shoot so conservatively that there is no recorded information in the far right one sixth of the histogram, you have effectively excluded half the possible information. So when you adjust levels in PS, you are stretching the 2048 tones you have actually recorded over the range which might have captured 4096 if you had exposed a little more to the right.

    As I say, it's probably old news to the people I so admire on this forum, but I feel like Archimedes getting out of the bath.

    I'm sure there are people who can comment more knowledgeably on this and correct me where I have not properly understood.
     
  2. gho

    gho

    Feb 7, 2005
    California
  3. Mike
    I had read a bit about this elsewhere and found that I did not (want to) believe it. I am finding that if you look into this carefully there is something to it and it IS worth knowing. Thanks for sharing this and now I got to go test it.
    Dave
     
  4. The camera is capable of capturing 4096 possible values per channel. 12 bits. But that is I guess irrelevant. The histogram goes from darkest to lightest.

    I don't understand why the instructor would indicate that there are more possible tones to the right side. Unless that is literally now the data is interpreted by the camera.

    Or does he mean the histogram isn't linear in the x axis? That is possible I suppose, though I have never read anything to indicate that. Certainly not in the manual or Thom's ebook.
     
    Last edited by a moderator: Sep 24, 2005
  5. nfoto

    nfoto Guest

    The instructor basically is correct. The first stop (in direction highlight to shadows) comprises 2048 values, the next 1024, then 512, 256,128,64,32,16,8,4,2 and that's the end of the number game. Now, this works out as 11 stops of range but in reality you cannot do much with a state having only two values (on or off), or even a few more, so the last stops of range before everything goes black are lost as well. In practice we end up with about 8 stops which is darned good compared to slide film. The main point here is to utilise the upper two stops of the dynamic range to maximise the tonal dynamics for the entire image. This means you have to "push" the exposure pretty close to clipping of the highlights, but still maintain details in the highlight areas. Since it is less likely that all three colour channels clip at the same point in the image, you can pull a surprising amount of detail from seemingly overexposed images by using the "highlight recovery" feature now becoming fashionable for raw converters.
     
  6. MontyDog

    MontyDog

    Jan 30, 2005
    #1064 - You have an error in your SQL syntax;
     
  7. JayR

    JayR

    385
    Jul 6, 2005
    Redmond, WA.
    Nice article. I think I understand the concept of exposing to the right but am not sure how to implement it.

    Do you just add exposure compensation to push the histogram right until it is all the way to the right w/o blowing highlights?

    What happens in the case where the light levels are really low and the ideal histogram should be to the left? For example, taking a skyline shot in the night. I would presume that most tonalities for such a picture would be to the left side of the histogram. What do you do in such cases?

    I appreciate the educational discussion as this is something that I would like to experiment with.

    Regards,
     
  8. Oh. Took it too literally.
     
  9. nfoto

    nfoto Guest

    Same thing. Push the histogram to the right if possible, adjust to taste in the raw-file converter. If however you shoot jpgs, thne the exposure must be "right" (or wrong, depending on your way of seeing things) in the first place, because you haven't got the necessary headroom to massage the image data after the exposure.
     
  10. heiko

    heiko

    May 15, 2005
    Israel
    Hello Bjorn,

    I never noticed the "highlight recovery" feature - which raw converters do support this feature?

    Sounds like an interesting approach that should work for many shots where only one channel is blown.

    Regards,
     
  11. nfoto

    nfoto Guest

    BibblePro 4.3a has it and it works surprisingly well. Still needs some manual tuning on difficult images, so it's not a feature you can switch on and use indiscriminately.
     
    Last edited by a moderator: Sep 24, 2005
  12. heiko

    heiko

    May 15, 2005
    Israel
    That's a quick response, Bjorn! What a pitty, I have NC and ACR :frown: at my disposal. Even if the highlight recovery takes some manual finetuning, it's certainly a cool feature.

    Thanks, Bjorn!
     
  13. Iliah

    Iliah

    Jan 29, 2005
    nowhere
    Digital cameras bring too much numbers to our life:)

    For now, I do not know of any commercial printer that can reliably reproduce details in more then 7 stops range, 6 2/3 is about maximum. Those stops correspond to RGB values from 242 to 33 in gamma=2.2 space, like Adobe RGB. Most printers can't reproduce details below RGB=45, 5 1/3 stops range. Shades of colour may be reproduced in print from RGB=250 to RGB=10, that is 10 stops. Prints evenly reproducing 10 stops look dull and flat. That is why converters do not apply simple gamma curve, but instead even more aggressively compress shadows and highlights, emphasizing contrast in mid-range of tones. Compression of shadows also helps preventing obvious noise.

    The "power of 2" distribution is to be discussed in linear gamma space, where in fact doubling the exposure doubles RGB values. But in gamma=2.2 space already the last stop contains less then 70 RGB levels (8-bit), or 1000 levels (in 12-bit space). S-shaped curves implemented in converters compress these further down, typically by half. On top of that, to make reasonable use of those levels we need to "compensate" exposure down in post-processing by about 1 stop.

    All this is not to undermine exposure to the right method, as we still want to make room for details and contrast for the stops following the first one.

    However,trying to keep specular highlights in the shot we actually expose to the left.

    Next, even division of histogram by f-stops is not correct, as the horizontal axis of the histogram is not linear. Less then 700(12-bit) levels are contained in the rghtmost 1/6 of the histogram.

    Also, it is often impossible to judge the exposure based on histogram, as histogram shows that something is clipped, but it does not show what exactly is clipped. Blinking highlights can be helpful here, but they are small and crude, do not take into account our capabilities of recovering them in postprocessing, and take time to watch them.
     
    Last edited by a moderator: Sep 24, 2005
  14. Chris101

    Chris101

    Feb 2, 2005
    Arizona
    Last edited by a moderator: Sep 25, 2005
  15. heiko

    heiko

    May 15, 2005
    Israel
  16. Same thing. Push the histogram to the right if possible, adjust to taste in the raw-file converter. If however you shoot jpgs, thne the exposure must be "right" (or wrong, depending on your way of seeing things) in the first place, because you haven't got the necessary headroom to massage the image data after the exposure.
    __________________
    Bjørn

    how does this affect combination files such as raw + basic? is it then neccessary to balance the exposure so that the jpeg is not blown?
     
  17. nfoto

    nfoto Guest

    You can always recreate a blown jpg from a healthy raw, so I wouldn't let the concern for the jpg override the exposure.
     
Loading...
Similar Threads Forum Date
A new recall on some Nikon EN-EL15 batteries General Technical Discussion Jul 7, 2017
Does spot meter exposure follow focus point? General Technical Discussion Apr 20, 2008
Follow focus on fast advancing subject. General Technical Discussion May 11, 2006