Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: marcmccalmont on July 19, 2006, 12:44:12 pm
-
When looking at a histogram either on camera (5D) or in PS2 is the histogram a simple linear scale or is it a logrithmic scale (stops) Thanks Marc
-
Linear, and the scale isn't in stops. It's arbitrary.
Michael
-
Linear, and the scale isn't in stops. It's arbitrary.
Michael
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=71150\")
Well, that depends on what is being plotted on the horizontal axis. The histogram is linear if one is plotting the pixel value of the gamma corrected image on the horizontal axis. If you want to relate the plotted value to the scene luminance, then the histogram is not linear but gamma corrected as explained in a post by Jonathin Wienke:
[a href=\"http://luminous-landscape.com/forum/index.php?showtopic=4723&pid=39033&mode=threaded&show=&st=&#entry39033]http://luminous-landscape.com/forum/index....st=&#entry39033[/url]
In a white paper on the Adobe site Bruce Fraser shows a linear histogram
http://www.adobe.com/digitalimag/pdfs/linear_gamma.pdf (http://www.adobe.com/digitalimag/pdfs/linear_gamma.pdf)
To make sense of the differences between linear, log, and gamma corrected, consider what is being plotted on the horizontal axis according to the following functions. In a linear histogram, the plotted value is linearly related to scene luminance. If the value plotted was according to log base 2, the plotted value would be in f/stops. If the gamma corrected value is being plotted, this represents a power function.
linear: f(x) = ax
log: f(x) = log(x)
power f(x) = a ^ x
-
interesting links. i noticed in C1 LE the Expsosure Evaluation on the capture tab is scaled at the bottom -4 -3 ... 0 where the distance between -4 / -3 is the smallest relative distance, -3 / -2 is a little longer ... -1 to 0 is the longest distance on the scale.
-
Then there's the interesting differences between color and luminance based Histograms:
http://www.cambridgeincolour.com/tutorials/histograms2.htm (http://www.cambridgeincolour.com/tutorials/histograms2.htm)
-
interesting links. i noticed in C1 LE the Expsosure Evaluation on the capture tab is scaled at the bottom -4 -3 ... 0 where the distance between -4 / -3 is the smallest relative distance, -3 / -2 is a little longer ... -1 to 0 is the longest distance on the scale.
[a href=\"index.php?act=findpost&pid=71195\"][{POST_SNAPBACK}][/a]
The Photoshop histogram does not have numbers on the x-axis, but basically goes from 0..255. How the histogram maps f/stops of luminance in the image on the histogram varies with the tone curve. Here is a Kodak Q-14 chart exposed with a D200 and converted with ACR with default settings except for a linear setting in the curves tab.
Three steps (0.3 density units) correspond to 1 f/stop, so you can see how the histogram varies with scene luminance. It is not quite a log histogram where the steps would be evenly spaced on the histogram. Spacing is widest in the middle.
[attachment=832:attachment]
-
bjanes: thanks
digitaldog: i remember reading that (several times)
i played around not too long ago with creating C M Y R G B squares and reading the RGB histogram vs. the luminosity. also converting to grayscale and observing the results. it was also fun to see the changes or no changes of applying curves or levels to those standard color patches.
-
Then there's the interesting differences between color and luminance based Histograms:
http://www.cambridgeincolour.com/tutorials/histograms2.htm (http://www.cambridgeincolour.com/tutorials/histograms2.htm)
[a href=\"index.php?act=findpost&pid=71197\"][{POST_SNAPBACK}][/a]
Yes that is an interesting difference. It makes you think that if a digital camera has only one histogram display, such as with the Nikon D70, it should be a composite RGB histogram such as Photoshop's default rather than a luminance histogram. With the former, no color channel will be blown if the histogram does not reach all the way to the right.
-
Thanks for all the input. I was hoping for an industry standard like octaves for sound measurement but I quess not. I did though learn a lot from the links and your input, I'll start "exposing to the right"-Marc