There have been dozens of articles and many videos published about the purpose and interpretation of the histogram in post-production. It serves a purpose in both the camera capture and post-production processes.
Tons of pulp has been generated about this tiny graph clarifying the general purpose of the histogram. But there is still much to be learned about this graph. This article will clarify some issues while focusing mainly on post-production concerns.
The individual RGB values are dominant in this image. Only the cracks in the sidewalk and spots of white in the cement cause the spikes on the outside edges of the graph.
Some articles I’ve read portray the histogram as the Holy Grail of digital photography and even describe an ideal histogram shape. Others describe the histogram as an accurate revelation of image health, sort of like a digital EKG. Still, others downplay the value of the statistical feedback completely and totally miss its primary purpose. Listed below are actual quotes from tutorials and articles about the histogram.
The two histograms reveal more information than you might think. The RGB graph reveals the same solid yellow and blue colors while the Luminosity graph represents the combined RGB luminosity (effectively, the grayscale equivalent values).
I’ve been adjusting images for decades; long before the graph was publicly introduced and I’ve worked with it every day since so I’d like to weigh-in on rumors and clarify some facts. Once you understand the histogram’s primary function and limitations, you’ll find it to be a solid feedback resource.
The Histogram’s Vertical Lines Defined
RUMOR #1: The histogram is a graphic blueprint of a digital image. The horizontal span represents the image’s tonal range and the vertical lines reflect the contrast of the image; the higher the vertical lines, the more contrast the picture contains.
FACT: The horizontal axis does reflect the image’s tonal range (from the darkest tones to the lightest), though the vertical lines reveal little about its contrast. Actually, the horizontal distribution is what reveals the overall contrast. Tones located mostly on the right reveal very light (or high-key) images while tones favoring the left side are darker (low-key) images.
The extreme right side wall represents white and the extreme left wall of the graph represents solid black. The highest (vertical) peak of the graph merely indicates the highest ratio of pixels containing that particular color tone as it relates to the others. The lowest vertical level on the graph indicates the tone color with the least number of pixels in the image.
The left graph represents a high-key or light toned image. The histogram on the right represents a low-key or dark toned image.
Histogram Shape
RUMOR #2: There is a preferred “mountain” shape for a histogram. The ideal shape displays a single peak beginning at the “ground” on one side, reaching upward into a bell shape near the middle, and tapering down to the ground on the other side. An ideal histogram contains information from all channels everywhere, from the left to the right in the graph.
FACT: There are as many histogram shapes as there are images. There is no such thing as a good or bad histogram and there is no such thing as an ideal histogram. Because these graphs reflect each image’s distribution of tones, you’ll be hard-pressed to find any two alike.
Only images containing mostly middle tones will actually display a true bell (mountain) shape on the histogram as in the one top left.
Clipping
RUMOR #3: The extreme left and extreme right sides of the histogram should never hit the “sidewalls” of the graph. If the left or right side hits and travels vertically up the wall, an undesirable effect called “clipping” will occur, indicating that either solid black or solid white “no-detail” areas will be visible in the image. Utilize the warning signs (The Blinkies and triangles) and avoid clipping on either tonal extremes of the graph.
FACT: Depending on whether the image is high-key (medium contrast on a pure white background) or low-key (dramatic lighting with a black background), either side of the mountain may actually resemble a tonal cliff. Real-life lighting dynamics make these wall-climbing graphs quite acceptable. Photos captured against white seamless backdrops are purposely exposed to produce dropout white backgrounds.
The U.S. Air Force Thunderbirds screaming across a very bright mid-summer South Florida sky were captured against totally white clouds. The apparent clipping shown in the histogram accurately reports the totally white clouds.
Tone Distribution
RUMOR #4: Histograms that display significant gaps on either side of the graph should be adjusted to distribute the tones more evenly. A well-shaped histogram is a happy histogram.
FACT: Real life lighting doesn’t demand that every scene contain both deep shadows and bright highlights. Images are sometimes brightened or darkened unnecessarily revealing a typical rookie editing mistake. Many times these bookend extremes establish an emotional mood that would be lost if the images were over-corrected in this manner.
This low-key shaded scene contains very little highlight shades at all.
Adjusting the image destroys the rich colors of the low-key scene.
Histogram Gaps
RUMOR #5: The silhouette of the histogram mountain should remain smooth, displaying no gaps or fissures in the mountain shape. These smooth tone transitions are necessary to maintain the photo’s full visual range. Gaps in the histogram’s silhouette indicate an interruption in the gradual tones and will result in banded or posterized stages. These gaps appear because the JPEG image is only 8-bit.
FACT: There are only 256 vertical bars presented in the Histogram. Each horizontal bar represents less than one-half of one percent (0.4%) of the total tonal range (100% / 256 = 0.390625%). Even if a photo contains a very gradual change in tones across a wide area (like an unclouded sky), your eyes will only perceive “banding” if the JPEG image has been degraded by repeated Save functions.
JPEG images contain a maximum of 256 levels (8-bits) of tone between black (solid color) and white (no color). Once JPEG files have been opened and saved a number of times, the number of tone levels can become significantly reduced and tone-banding may occur.
These two histograms appear quite different though the images are very similar. As you can see, “gappy” pictures rarely interrupt the flowing tones in a photo and often clarify more detail.
Full Range Images
RUMOR #6: 8-bit images (256 levels of tone/color per Grayscale/RGB channel) are required for an image to display the full range of detail contained in a high-resolution digital image.
FACT: The human eye is designed to focus on detail in a scene or image. Detail is a product of contrast, and contrast is only noticeable when adjacent colors display significant differences. For the most part, the fewer colors that are displayed, the more obvious are the differences in those colors. This sounds bad but it is actually quite beneficial. It may fly in the face of popular logic but there is a fundamental truth to be recognized. The picture with the least colors (also known as bit depth) is many times the most detailed picture.
The top image contains 256 levels of color per RGB channel; or up to 16,800,000,000 colors. The bottom image contains only 15 levels of color per RGB channel; only 3375 possible colors. This image uses less than 5% of the tones that are used in the top image. Don’t live or die on the issue of bit depth. It is important, but like other issues in life, more isn’t necessarily better.
There is, of course, a reasonable limit to this reduced color observation. Too few levels of color will lose the smooth transitions between colors and thus will also lose detail.
Try this exercise: open a full-range photo in Photoshop and duplicate the image on another layer. Open the Histogram (Window/Histogram). Now select Image > Adjustments > Posterize, enter the number 15 and view the image as a Preview. Watch the Histogram window as you preview the image. The graph will display only 15 vertical columns instead of 256 but the image will look pretty much the same.
The point I’m making is that “gaps” in a normally smooth histogram doesn’t necessarily indicate a visual disaster. Instead, a few gaps might just inject a little more drama in your images.
Smooth Flowing Tones
RUMOR #7: Digital captures should display the highest number of tones possible in order to preserve the smooth continuous tones.
FACT: While nature provides an extreme level of dynamic range and true continuous tone gradation, there is no such thing as “continuous tone” digital images. The word “digital” affirms this statement. Virtually all digital images are comprised of square pixels displaying individual tone values. The perception of continuous tone is an illusion.
These blue pixels were captured from the sky of the Thunderbirds image above. The patch of sky appears as continuous tone in the original picture although the enlarged individual pixels vary quite a bit in value. The top histogram shows the combined channel luminance while the bottom shows individual RGB channel values. It’s complicated.
Exhaustive Reference
RUMOR #8: A Histogram is an exhaustive systematic and statistical accounting of all the internal colors and tones of an image, from dark to light (100% – 0%).
FACT: Each histogram does reveal the relative placement and distribution of all tones and colors, but due to its size, there is a serious limitation to its accuracy. Since editing software histograms are based on a horizontal graph only 256 pixels wide, each representation is a basic overview at best. If the full range of possible colors were truly represented by a single graph, the chart would occupy the wall of a good size room!
Both the top and bottom tonal adjustments were severe enough to create significant gaps in the histogram and to alter the shift of tones in each range, though there are no visible interruptions in the smoothness of each gradation.
Let me break down the numbers. This 256 pixel-wide graph portrays each image’s potential color range using an 8-bit (256 level) interpolation. This means that all 16.8 million possible colors are represented in a mere 256 horizontal point histogram. Tones change levels in 0.4% increments. The graph significantly exaggerates the difference between minor shifts in tonal value.
Human eyes barely perceive a half-percent (0.5%) difference between tones, which is why 256 levels in a JPEG image provides the illusion of continuous tone. This means the histogram uses less than two vertical columns to represent a single percent of change in value.
The two bottom patches are separated by only 1% in tone, though your eyes may have trouble seeing the difference. The top band shows a gradual change between the two values.
What does all this mean? Quite simply, the histogram delivers a good estimation of overall tone distribution but cannot be relied on for accurate measurement. A few gaps in the graph will rarely be visible to the human eye.
Wrap Up
The histogram is a valuable instrument intended to deliver a quick overview of the makeup and tonal shape of digital images. It was never intended to be a scary reference tool.
Learn to use the histogram to deliver interactive feedback as you work through your editing steps but appreciate the graph for the information it provides; mostly defining the distribution of tones, highlights, and shadows.
What the histogram does not show is whether the image requires internal adjustments to reveal hidden detail. That’s another topic altogether.
Push pixels around and stay focused.
The post Shedding Light on the Histogram – 8 Rumors and the Real Facts About this Graph appeared first on Digital Photography School.
Digital Photography School
You must be logged in to post a comment.