Using the Histogram as an exposure guide

tomtom2245
12 Mar, 2017 21:56
Still trying to learn some of the basics but how effective is using the histogram at determining exposure length? Some of my early attempts have the histogram peak going above the max "y" value on the graph, hence maxing out that region. I have noticed that those images that do max out part of the histogram are much harder to post process. I'm going to assume that all exposures should avoid maxing out any section of the histogram and if you happen to do that while shooting different exposure lengths, you should back off the exposure or lower the ISO. Any help is appreciated!
tonyhallas
13 Mar, 2017 03:33
Hi Tom,

   Are you using a DSLR or a CCD?  If it is a DSLR I have found that exposing with the histogram towards the center yields good results … and you are correct, if it is smack up against a boundary that is no good. It's called "clipping" and means you have no data. It's either lost in the absolute black, or lost in the white … depending on where you are.  Histograms have one more useful function … they can show in post processing when you have stretched too far. If the histogram "combs", ie, turns from a smooth graph into a bunch of lines, you have gone too far and need to back off.

    Hope this helps,

       Tony
tomtom2245
13 Mar, 2017 08:55
tonyhallas
Hi Tom,   Are you using a DSLR or a CCD?  If it is a DSLR I have found that exposing with the histogram towards the center yields good results … and you are correct, if it is smack up against a boundary that is no good. It's called "clipping" and means you have no data. It's either lost in the absolute black, or lost in the white … depending on where you are.  Histograms have one more useful function … they can show in post processing when you have stretched too far. If the histogram "combs", ie, turns from a smooth graph into a bunch of lines, you have gone too far and need to back off.

    Hope this helps,

       Tony
I'm using a DSLR and I'm more concerned with it hitting the top of the histogram like it is maxing out a particular tonal value if that makes sense.
tonyhallas
13 Mar, 2017 14:50
Hi Tom,

   Not sure what you are imaging but under a dark sky it's pretty hard to "burn it up" on the right side of the histogram. Maybe the center of M42, or the moon, but if shooting the Milky Way under a dark sky having the histogram somewhere in the middle will be plenty of signal. One of the problems going too far to the right is that you might be recording some light pollution that will drown out your data. Typically under a dark sky it is actually pretty hard to push the histogram all the way to the right.
   Best,
       Tony
bobzeq25
15 Mar, 2017 02:48
Here's how to get it right.  There'll be a prominent "skyfog peak".  You want that about 1/3 over from the left on the subframes.  It will never max out on the y-axis, even in light polluted skies, if it did your frame would be pure white (it should still be surprisingly bright).  Put it about 1/3 over, you never even have to think about maxing it out.

The stack will always be dismayingly dark, if you do it properly.  You brighten it in your processing program.  NOT in DSS.

Some stars will always max out.  Unavoidable.  Maybe some galaxy centers or other parts of the image.
Edited 15 Mar, 2017 02:53
MarcovanderKooij
15 Mar, 2017 02:51
tonyhallas
Hi Tom,   Not sure what you are imaging but under a dark sky it's pretty hard to "burn it up" on the right side of the histogram. Maybe the center of M42, or the moon, but if shooting the Milky Way under a dark sky having the histogram somewhere in the middle will be plenty of signal. One of the problems going too far to the right is that you might be recording some light pollution that will drown out your data. Typically under a dark sky it is actually pretty hard to push the histogram all the way to the right.
   Best,
       Tony
MarcovanderKooij
15 Mar, 2017 02:57
Hi.. Tony and all…
Isn't it better to keep the signal a bit further to the left in the histogram (not mid way)? That way we utilize better the dynamic range of the sensor while not unnecessarily causing saturation on the right for bright stars.
As long as we dont get quantization noise for valuable dark information we would be good.  I like to keep dark background at around .05 or so. What do you think?
Marco
tomtom2245
15 Mar, 2017 03:09
So here is an example of the type of histogram I am referring to, just a sample one I came across via Google but very similar to what some of mine look like. The peak as you can see is on the left hand side about 1/4 of the way over but yet the peak maxes out at the top of the histogram.

MarcovanderKooij
15 Mar, 2017 03:34
Yes. So the peak represents the majority of pixels. That must be the backgrpund at around 20 to 25 precent of maximum.  If you load the image in pixinsight or other software what are the values of background?
So i am still wondering if it is better to keep background at lower value say t or 10 percent.
tonyhallas
15 Mar, 2017 03:57
Hi Tom,

   The top of the histogram only shows HOW MANY of a particular exposure value you have … the "light value" … how light or dark your values are, goes left to right with extreme left being pure black and extreme right being white.  I have taken a lot of DSLR images and I used to think that if I had the columns in the position of your illustration I would be OK.  But then working with the data I saw that pushing the columns further to the right via more exposure or faster f/ratio gave me better signal to noise.  IMHO you want them more towards the center ( left to right ).

    Best,

         Tony
tomtom2245
15 Mar, 2017 03:57
Marco van der Kooij
Yes. So the peak represents the majority of pixels. That must be the backgrpund at around 20 to 25 precent of maximum.  If you load the image in pixinsight or other software what are the values of background?
So i am still wondering if it is better to keep background at lower value say t or 10 percent.
I'm away from my home computer right now so I'll have to get you those values later but what exactly are you looking for? I use Photoshop CC so I could use the eyedropper tool and show you where it falls on the histogram.
bobzeq25
21 Mar, 2017 01:47
The vertical axis has no units listed, it's probably just too small.  That histogram is about what you're looking for.  A touch too far right (only 20% over if the x axis scale tops out at 100%), not enough to matter.
Edited 21 Mar, 2017 01:49
bobzeq25
21 Mar, 2017 01:55
Marco van der Kooij
Yes. So the peak represents the majority of pixels. That must be the backgrpund at around 20 to 25 precent of maximum.  If you load the image in pixinsight or other software what are the values of background?So i am still wondering if it is better to keep background at lower value say t or 10 percent.
No, assuming the histogram is stretched (and not linear).  The one above It looks like it.  You want to be careful not to clip the low end.  It's helpful to be able to adjust the black point to get the background as desired.  The amount of dynamic range you're losing is not significant.

If you measure the background in PI it's a linear (unstretched) value.  That will be smaller.  10% of full scale is fine, I generally run at 5% or even a bit lower.  It depends on your precise setup.   And whether it's an unprocessed light, or calibarated.
Edited 24 Mar, 2017 21:03
swordfish
22 Mar, 2017 15:44
bobzeq25
The vertical axis has no units listed, it's probably just too small.  That histogram is about what you're looking for.  A touch too far right (only 20% over if the x axis scale tops out at 100%), not enough to matter.
This is 100% correct. There is no such thing as vertical clipping, it just typically (and annoyingly) goes off the bounds of the histogram. I'm a newbie at this as well but I believe the increase in difficulty in post you are referring to comes from how narrow the histogram is. If most of your data falls within the same range of values and it is a smaller range, it limits your flexibility. You can try adjusting your levels to widen the slopes of your histogram a bit more.
 
Register or login to create to post a reply.