The Histogram is a statistical tool for graphically looking at the data contained in a bitmapped graphic file. All it does is tabulate the pixels by level (any of 256 in Photoshop) and displays them as a frequency distribution graph (called histogram). The histogram does not show you the picture or any representation of it other than pure math. It's just a statistical tool to tell you what kinds of densities and how many of them at each density level are found in the picture.
Shown here is a histogram of a digital picture (Sony FD91). It shows the number of pixels at each of the 256 image levels (move your cursor to the level you want to see the total for). In this case, level 128 (exactly 50%) is 3,785. Since there are 786,432 pixels in the entire image, .4% of them are at exactly 50% gray. By moving the mouse point above the gray scale you can tell how many
But look at the ends. Notice that there isn't much height on our histogram. This means there are no blacks (on the left) or whites (on the right) in this picture. Since there is no line on the extreme left or right there are no solid black areas or washed out whites. Neither solid black or pure white carry any apparent image and therefore the histogram may alert you to this problem. The pixels in this picture are far into the midtones.
What color is this picture? The above histogram gives us that average of the three colors, the gray equivalent of the picture. In newer Photoshops it is referred to as Luminescence. But our picture is in color and we should take a look at the individual color histograms to learn more about our picture.
The Red and Green channels show no particular problem, the blue shows a lot of very heavy shadow detail. But why doesn't this show up on the Gray histogram? Each of the color channels only define a part of a pixel. Even if the blue component is solid, the other two channels may be light bring the pixel, on the average, to a lighter gray value.
The Picture
But what does the histogram tell us about our picture? We can't tell by the historgram. I t is merely a statistical waly of looking at the data. There is no way of knowing what the picture is or even how
By the way, here's the picture. This is not the version showing the above histrogram, this picture is image sized to fit this web page and may be slightly different. It obviously contains a lot less than three quarters of a million pixels!
It is a picture of a Blueberry, which should account for the heavily dark blue channel. There is no white and no black on the picture. It is actually all green midtones, just as the Gray histogram shows. Pictures do not have to have a black area and a white area to be an acceptable picture. The densities represented on the historgram do not dictate the picture, it is the picture that merely contains pixels of various densities that make up the picture. The histogram merely reports what is in the picture. Each picture should report a different histogram because each picture is different.
Detecting a bad picture with the Histogram
Here is a picture that shows a problem. When there are a large number of pixels that are white they cannot carry any apparent image. They are pure white. If there are only a few of them and they are scattered around the image this is not be a serious problem. It would be an indication of a very light highlight. But when there are a lot of them; 15,755 in this picture, it is an indication that an area that may be solid white carrying no apparent imge. This is historically called "washed-out", where highlights hold no picture detail. There is nothing wrong with having some whites but here is an example of too many.
Looking at the shadow, we find that there are 251 solid black pixels. Not such a bad problem because black does produce those dark shadows and adds to the contrast of the picture. But the first four levels of black total 17,370 and this may indicate that there is little near shadow detail. There is little to distinguish between black and almost black. They may all look about the same. By the way, there are 29,209 pixels in the four lightest levels. Not much for near highlight detail either!
And here is the unretouched picture, a Beaver Dam. A flare in the sky destroys all the near highlight detail. We see that the shadows are very, very dark holding almost no detail. In all, it is not such a good picture. We sort of can see that, but the Histogram tells us exactly why it is not right. This is probably a picture that should be discarded unless you want to spend a lot of time working on it. In the field, use your LCD screen and preview. If you notice a problem like this, immediately delete the image and try again with different exposure settings.
According to the Histogram the shadows are too dark. We therefore have to improve the near shadows. The flare in the sky is very annoying. We should try and minimize it by increasing the near highlight details. As a result of the problems reported by the Hostograms, we have decided not to work on this picture. Maybe someday we will return to the Beaver Dam and take a better picture, one that shows a better range of tones, no washout and more near shadow detail. We won't easily be able to get it out of this one!
Adjusting the picture changes the Histogram
When you make changes to your picture you are changing the Histogram. Once finished, the picture should show an improved picture. Let's see the Blueberry picture improved a bit. At left, the berry is now blue, maybe even a bit too blue. But it does get the point across that it is a blueberry! No sharpening was done to this picture because the leaf at right showed signs of oversharpening. Perhaps a local sharpening of the blueberry itself might help. Perhaps too much effort for only a slight bit of improvement.
The histograms? they've changed a bit from the beginning and now show a more uniform distribution of densities. Spreading the tones through the scale allows for more representation from highlights, midtones and shadows. A fuller range of tones, especially in the colors produces a more lifelike picture. Notice that there are tones in the gray extending to about the edges of the scale on both ends.
According to the Histogram we have the makings of a much better picture, so do you like it? Again, the Histogram can be of no value. How we like pictures is purely subjective having nothing to do with technical details or requirements. In the end, we are left with only our better judgement. If it looks good, it probably is regardless of what the Histogram tells us.
Conclusion
So what do we do with the Histogram? Simply use it as a guide to the technical evaluation of our picture. Use it as a way of telling if the picture is good or not. Most important, keep in mind the technical information it presents us while we merrily go about making all sorts of adjustments. Look at the Histogram after you make an adjustment such as Sharpen or Curves and see if you aren't doing some technical damage in the quest for aesthetic perfection.
Over time you will learn to recognize potential problems with your images by studying the Histogram. Each image will present its own Histogram and that will reflect the uniqueness of that image. There is no answer to the Histogram. There is no formula to follow and no set of procedures to follow if you see a certain Histogram. You are going to have to see a lot of histograms from a lot of pictures before you get the message that you are looking into the inner workings of a particular unique picture.
The Histogram will tell you the technical details of your picture in much the same way your screen view shows you a graphical representation of the data, which coincidentally looks just like the picture. The Histogram does not look like the picture but is still the same data. Learn to look at the Histogram in the same way you look at the screen view.
