Same Brightness Different Contrast
Why average pixel intensity cannot describe how an image looks
Two grayscale images can have exactly the same average brightness and still look completely different. The reason is simple: an average tells us where the centre of the pixel values lies, but contrast depends on how widely those values are spread.
The question
Suppose Image A contains mostly medium gray pixels between 110 and 140, while Image B contains values spread from 20 to 230. Both images have the same average intensity. Which image is likely to have higher contrast?
Image B is the better answer. Its wider intensity range allows much darker and much brighter regions to appear in the same image. Image A stays close to mid-gray, so the visual differences between its pixels are smaller.
Figure 1 Two images can have similar means but very different intensity spreads
Brightness and contrast describe different things
Brightness describes the overall lightness or darkness of an image. In a grayscale image, it is often summarized using the mean pixel intensity. Contrast describes the difference between darker and brighter parts of the image. It depends on the distribution of pixel values, not only on their average.
This distinction matters because the mean compresses an entire image into one number. During that compression, it loses information about the distances between pixel values. Many different distributions can share the same mean.
A small numerical example
Both sets have a mean of 125. Yet Image B contains values much farther from 125. That larger separation creates stronger differences between light and dark pixels, which we perceive as higher contrast.
What the histogram reveals
An image histogram counts how many pixels occur at each intensity. For an 8-bit grayscale image, the horizontal axis usually runs from 0 for black to 255 for white. The vertical axis shows how frequently each intensity occurs.
A narrow histogram concentrated near the middle usually indicates low contrast. A histogram spread across a wider range usually indicates stronger tonal variation. The histogram therefore reveals information that the mean cannot.
Standard deviation gives the spread a number
Get Asha Natty’s stories in your inbox
Join Medium for free to get updates from this writer.
Standard deviation measures how far pixel intensities typically lie from the mean. For N pixels with intensities I and mean intensity mu, it is calculated as:
A small standard deviation means most pixels stay close to the mean. A large standard deviation means the intensities are more widely dispersed. For many grayscale images, this makes standard deviation a useful global indicator of contrast, although it does not describe every aspect of perceived contrast.
The logic behind the question
The question is designed to test whether we confuse a measure of centre with a measure of variation. Average intensity answers, roughly, how bright the image is overall. It does not answer how strongly the pixels differ from one another.
A reliable reasoning process is:
Identify what the mean measures: the centre of the intensity values.
Look at the range or distribution: narrow values suggest lower contrast, while a wider spread suggests higher contrast.
Do not conclude that equal means imply equal images. The same average can come from many different pixel arrangements and distributions.
What this question does not tell us
A histogram ignores spatial arrangement. Two images can have identical histograms while placing the same gray values in completely different locations. The histogram can describe the intensity distribution, but it cannot reveal which objects are present, where edges occur, or how the pixels form shapes.
Standard deviation is also a global measure. An image may have high global contrast because of a few extreme regions while still having weak local contrast in an important area. Local contrast measures and spatial analysis are needed when the location of detail matters.
Why the distinction matters in computer vision
Brightness and contrast affect thresholding, edge detection, feature extraction, recognition, and segmentation. A model trained mainly on well-lit images may struggle when the same objects appear under flat lighting or strong shadows. Contrast enhancement can reveal useful structure, but careless enhancement can also amplify noise or remove subtle tonal relationships.
The practical lesson is to inspect more than one summary statistic. Mean intensity, histogram shape, standard deviation, and spatial structure each describe a different part of the image. Together they provide a more complete picture than any single number.
Final takeaway
Mean tells us about average brightness. Spread tells us about contrast. Two images can share the same mean intensity while having very different histograms, standard deviations, and visual appearance. That is the central concept the question is testing.
Further reading
Robert Keim Understanding Contrast Histograms and Standard Deviation in Digital Imagery
Adobe Help Center View histograms and pixel values in Photoshop
