I admit it. I need help in understanding how histograms work.
I try to use the histogram graphs on my Nikon D80 whenever I shoot. With the histogram graphs turned on I see on the back of my camera a small thumbnail of the photograph and four histogram charts, white, red, green, and blue. I do not know what each of them mean. I try to see how they relate to the image but I am not getting it. Sometimes I see a series of tall peaks, other times a single peak, and the occasional hills that reach only half way to the top. I am looking at one now where the single peak is near the left which I think should indicate a darker exposure but the image is not dark at all.
Inspite of my lack of understanding, most of my images turn out well.
Enough rambling. I am turning to you for help.
I appreciate all the responses to my question. Thanks for your help.
A histogram relates the distribution of data across a sample (your photo). In this case, it is broken into 256 buckets indicating the value of the pixels from 0 to 255. Each time a pixel lands in the bucket, it makes the pile in that bucket higher. So lets say a pixel holds a brightness value of 198, it drops into the 198 bucket.
Now there are three colors and white. White represents the luminosity. I suppose its equal to R+G+B/3. Red represents the distribution of red pixel values, Green = green pixels and blue = blue pixels.
A histogram with two peaks, one on the left ad one on the right indicates a high contrast scene where few pixels have ended up in mid tones. A histogram shaped like a big hill centered on the middle indicates that the distribution of pixels is largely around the middle tones.
I find the white histogram to be the most useful. It gives you instant feedback regarding your exposure. The right side of the histogram represents highlights and the left side is shadows. If the graph of the histogram touches the right edge of the frame, it means there are blown out highlights. If the left edge of the graph touches the frame it means that there are blocked up shadows. The old saying " expose for the highlights" has now become "whites to the right." It means that you should set your exposure so that the graph moves as far as possible to the right of the frame without actually touching the edge of the frame.
The histogram, in conjunction with the highlight warning (the blinkies), completely takes the guesswork out of exposure. You know instantly if you have blown the exposure and can adjust and re-shoot if you need to.
I find the red, blue, and green histograms to be less useful in the tiny confines of the camera LCD. You should be able to use your menu to show the white histogram only.
Will - Check out this collection of tutorials from Cambridge in Colour. I think they are one of the best overall references for this type of thing. There are two on histograms in this case, about a third of the way down, but they are all good.
Posted by Paul DeCesare on June 17, 2009 at 11:04 PM
David, what a great collection of tutorials... you win the pirate hat for tonight's submission! (And as soon as I figure out what a pirate hat has to do with digital photography, I'll post the answer).
William, if you want to take this whole subject even further, this particular book was a huge help a couple years back when I was trying to figure out the histogram and how it related to my RAW files once in Photoshop (Elements at that time). Once you understand the histogram, you'll then want to know what to adjust BASED on that histogram once you edit the image. The first thing I had to learn was that the shape of the histogram is irrelevant 99% of the time.
I agree with Paul... don't worry at all about shapes... keep your eyes focused on either side... idealy you'd want a bell curve feathering perfectly out from the middle... but that's not reality.... just keep whatever shape you have near the middle (or slightly towards the high end... I read that somewhere)... best of luck....post another question if we haven't helped!! Another thing about photography....you're in the best place to learn...what you need to know (and don't need to know). Everyone wants to help, but everyone has their own "language". For instance, I can't stand the tech speak that is often involved with computers and/or digital photography. Bottom line: keep the bars of the histogram between the lines, but worry more about perspective, angles, and light.
Posted by David Taylor on June 18, 2009 at 1:01 AM
Definitely agree with Todd and Paul - the shapes are inconsequential, where they fall on the graph are most important. BUT, remember that the histogram is NOT generated off of the RAW data file. It is an approximation made off the in camera JPEG. Even if you don't shoot RAW + JPEG, the camera still generates one for thumbnail representation "in camera" viewing. That's why you can still turn on things like "picture styles" in the camera and the LCD will show you saturation/contrast/sharpness adjustments. These "styles" are shown when you view the image on your camera LCD, even though they are not "burned in" to your final RAW file.
So, make sure you learn how your LCD histogram corresponds to your post production software's histogram (ie: PS CS3/4, Elements, Bridge, LR, camera RAW converters, etc.). You won't know how accurate your camera LCD is until you compare it with your software histogram. Then you can make more informed decisions about exposure "in the field".
Hope this helps.
Very good Dave.... let me tie that into raw files and conversion.... I think the best lesson can be learned when one goes to find what you can "bring back" on conversion.... I don't think that can be taught... but I do know if you're near the middle, there is nothing to bring back. I do think we're getting beyond the scope.... it's almost what we need to do is teach feel... I don't even trust mine!!! William.... look up on this thread for wisdom, than listen to Dave and myself.
Keep in mind that a histogram is generated by the software, on the particular device you're viewing the photo. Based on the numeric values in the image, the histogram will look different on your camera, computer, etc. Here's a link to an article I wrote a while back, about this... hope it is helpful:
Will,
This is what I was talking about earlier. Everybody needs some technical assistance sometime. EVERYBODY. I recognize this dude's name and is legit. He can definitely split up your histogram into whatever. I was trying to approach it from a hands on, artistic point of view
Posted by Paul DeCesare on June 18, 2009 at 10:12 AM
Nat, great article on Histograms!
In fact, you have a very well-done site overall. I just signed up for your newsletter.
I am certainly getting plenty of help in understanding histograms. For that I am grateful. I really appreciate you stepping up to help me out. Nat, I found your article well written and printed it. I am very interested in your book too.
What throws me off is the shape of the histograms. Would I be able to anticipate what shape the correct histogram should be before I take the shot? Or does that really matter as long as I avoid getting too much up against the left or the right?
None of us used a histogram to determine the correct exposure when we used film cameras. Through practice we learned how to get the correct exposure with reasonable consistancy. Now we carry over that knowledge and skill to the digital camera. So do we need to rely on the histogram for the correct exposure or is it merely to verify that we got it right? I use the histogram to verify that I got it right.
Thank you very much for coming to my aid. I am still busy soaking all the information in, especially when it comes to applying it to my own shooting.
Or does that really matter as long as I avoid getting too much up against the left or the right?
That's the aim when using the in-camera histogram.
See this essay titled 'Expose to the Right' for some good info. on why keeping the information on the histogram as far to the right as possible without clipping (overexposing) pixels is important.
The on-camera histogram is one of the best features that digital cameras possess, IMO.
Posted by David Taylor on June 19, 2009 at 6:14 AM
The shape of the histogram itself really doesn't matter - it is just a graphical representation of where the colors (and their luminosity) fall through the range of the exposure. I always display the RGB + Luminosity histogram on my LCD. A good example is photographing a columbine - the red of the flower can be quite intense. If you look at the "R" channel on your camera, it may be clipped on the RED highlights, so I might lower my exposure to compensate.
Every scene you photograph will have a different histogram shape, so there is no "correct" shape. You won't always get a bell shaped curve.
Posted by Nat Coalson on June 19, 2009 at 11:28 AM
Good stuff here. I agree that using the histogram while shooting is "best-practice".... it takes some time, though, to learn how that translates in your raw processor. Just remember that what you see on the camera (preview and histogram) is based on its internal processing (same as if you captured a JPG). You might see clipping on your camera histogram and not in your computer...
If you see spikes at either end, the histogram is telling you there are pixels of either solid black (left) or pure white (right). A little clipping in shadows is usually not a problem. Clipped highlights should be avoided if possible, but when you have hot spots (like specular reflections) it can't be avoided. Expose for the most important parts of the scene.
The RGB separations are only minimally helpful; what you really care about is the brighness/luminance histogram, because it deals exclusively with tone, not color. If you see clipping in only one or two channels, don't worry about it.
Don't try to use the shape of the histogram to decide if the exposure is "correct"; what matters most is how much data you're capturing and the data is distributed over the scale. You can correlate this to the captured frame (as described in my article).
Finally, John's point about exposing to the right is important. A digital capture is linear which, put simply, means that more light=more data. If I recall the numbers correctly, 75% of all data in a digital capture lies in the top 25% of tones... (it might even be 90/10?... I forget)
In other words, dark areas of the capture have MUCH less data than bright ones. For this reason, when processing it's easier to manipulate lighter captures (including darkening them) than it is to adjust darker captures.... it all comes down to the amount of data that is captured.
So if you aren't certain to nail the capture dead-on, it's best to err on the side of a slight overexposure.
Whenever possible, avoid underexposing any digital capture, unless there's a specific reason for it.
To add to what Nat and John said....that Luminous Landscape article above doesn't really discuss it although it mentions it several times. In addition to more levels available in the brighter tones, that's also where the highest signal-to-noise ratio exists. So exposing-to-the-right's (ETTR's) biggest advantage in my opinion is in creating the least noise in the image. So overexposing and then correcting during RAW conversion has two benefits in the shadows/darker tones :
1) More potential values captured, enabling better manipulations during processing with less posterization issues
2) Less noise in the darker areas of the image (in the whole image, really, but shadows show it much much more than highlights). I think this one is the bigger benefit.
But as the article says - if, for your shooting and technical criteria, you don't see a wirthwhile improvement, then by all means shoot however you normally would. Whatever works for you!
And to go a little off topic...some people will take an exposure that is several stops overexposed just to use that one for the shadow areas. And a "normal" exposure for the midtones and highlights, then blend the two images after the fact. I've done it on a few occasions, and it can significantly reduce noise. The blending isn't simple and may require some other software though, unless the shadow areas are very well segregated from the rest of the image.
Discussion
15 repliesI try to use the histogram graphs on my Nikon D80 whenever I shoot. With the histogram graphs turned on I see on the back of my camera a small thumbnail of the photograph and four histogram charts, white, red, green, and blue. I do not know what each of them mean. I try to see how they relate to the image but I am not getting it. Sometimes I see a series of tall peaks, other times a single peak, and the occasional hills that reach only half way to the top. I am looking at one now where the single peak is near the left which I think should indicate a darker exposure but the image is not dark at all.
Inspite of my lack of understanding, most of my images turn out well.
Enough rambling. I am turning to you for help.
I appreciate all the responses to my question. Thanks for your help.
Will
Edmonton, Alberta
Canada
Now there are three colors and white. White represents the luminosity. I suppose its equal to R+G+B/3. Red represents the distribution of red pixel values, Green = green pixels and blue = blue pixels.
A histogram with two peaks, one on the left ad one on the right indicates a high contrast scene where few pixels have ended up in mid tones. A histogram shaped like a big hill centered on the middle indicates that the distribution of pixels is largely around the middle tones.
Hope that helps.
The histogram, in conjunction with the highlight warning (the blinkies), completely takes the guesswork out of exposure. You know instantly if you have blown the exposure and can adjust and re-shoot if you need to.
I find the red, blue, and green histograms to be less useful in the tiny confines of the camera LCD. You should be able to use your menu to show the white histogram only.
William, if you want to take this whole subject even further, this particular book was a huge help a couple years back when I was trying to figure out the histogram and how it related to my RAW files once in Photoshop (Elements at that time). Once you understand the histogram, you'll then want to know what to adjust BASED on that histogram once you edit the image. The first thing I had to learn was that the shape of the histogram is irrelevant 99% of the time.
So, make sure you learn how your LCD histogram corresponds to your post production software's histogram (ie: PS CS3/4, Elements, Bridge, LR, camera RAW converters, etc.). You won't know how accurate your camera LCD is until you compare it with your software histogram. Then you can make more informed decisions about exposure "in the field".
Hope this helps.
http://bit.ly/histograms-1
More about this is in my Lightroom book my Lightroom book
This is what I was talking about earlier. Everybody needs some technical assistance sometime. EVERYBODY. I recognize this dude's name and is legit. He can definitely split up your histogram into whatever. I was trying to approach it from a hands on, artistic point of view
In fact, you have a very well-done site overall. I just signed up for your newsletter.
What throws me off is the shape of the histograms. Would I be able to anticipate what shape the correct histogram should be before I take the shot? Or does that really matter as long as I avoid getting too much up against the left or the right?
None of us used a histogram to determine the correct exposure when we used film cameras. Through practice we learned how to get the correct exposure with reasonable consistancy. Now we carry over that knowledge and skill to the digital camera. So do we need to rely on the histogram for the correct exposure or is it merely to verify that we got it right? I use the histogram to verify that I got it right.
Thank you very much for coming to my aid. I am still busy soaking all the information in, especially when it comes to applying it to my own shooting.
Will
Edmonton, Alberta
Canada
That's the aim when using the in-camera histogram.
See this essay titled 'Expose to the Right' for some good info. on why keeping the information on the histogram as far to the right as possible without clipping (overexposing) pixels is important.
The on-camera histogram is one of the best features that digital cameras possess, IMO.
Every scene you photograph will have a different histogram shape, so there is no "correct" shape. You won't always get a bell shaped curve.
If you see spikes at either end, the histogram is telling you there are pixels of either solid black (left) or pure white (right). A little clipping in shadows is usually not a problem. Clipped highlights should be avoided if possible, but when you have hot spots (like specular reflections) it can't be avoided. Expose for the most important parts of the scene.
The RGB separations are only minimally helpful; what you really care about is the brighness/luminance histogram, because it deals exclusively with tone, not color. If you see clipping in only one or two channels, don't worry about it.
Don't try to use the shape of the histogram to decide if the exposure is "correct"; what matters most is how much data you're capturing and the data is distributed over the scale. You can correlate this to the captured frame (as described in my article).
Finally, John's point about exposing to the right is important. A digital capture is linear which, put simply, means that more light=more data. If I recall the numbers correctly, 75% of all data in a digital capture lies in the top 25% of tones... (it might even be 90/10?... I forget)
In other words, dark areas of the capture have MUCH less data than bright ones. For this reason, when processing it's easier to manipulate lighter captures (including darkening them) than it is to adjust darker captures.... it all comes down to the amount of data that is captured.
So if you aren't certain to nail the capture dead-on, it's best to err on the side of a slight overexposure.
Whenever possible, avoid underexposing any digital capture, unless there's a specific reason for it.
1) More potential values captured, enabling better manipulations during processing with less posterization issues
2) Less noise in the darker areas of the image (in the whole image, really, but shadows show it much much more than highlights). I think this one is the bigger benefit.
But as the article says - if, for your shooting and technical criteria, you don't see a wirthwhile improvement, then by all means shoot however you normally would. Whatever works for you!
And to go a little off topic...some people will take an exposure that is several stops overexposed just to use that one for the shadow areas. And a "normal" exposure for the midtones and highlights, then blend the two images after the fact. I've done it on a few occasions, and it can significantly reduce noise. The blending isn't simple and may require some other software though, unless the shadow areas are very well segregated from the rest of the image.