Does higher resolution eliminate banding?

No. Resolution controls spatial sampling, while banding often concerns the number and distribution of tonal levels.

Does dithering add more colors to an 8-bit channel?

It does not increase the channel's possible codes. It distributes quantization error spatially so transitions can look smoother.

Can a high-bit-depth source still look banded?

Yes. Later processing, output conversion, compression or the display path may reduce effective precision or reveal steps.

AI Glossary / Computer vision and multimodal AI / Beginner

Banding

Definition

Banding is the appearance of visible steps or stripes in a region that should change smoothly in brightness or color. In digital imaging, it often occurs when too few distinct values remain to represent a gradual transition, such as a sky or a soft gradient.

What it means

A smooth gradient should change by small amounts across the image. If many neighboring pixels are forced to share one value and then jump together to the next, the eye may see boundaries between broad tonal regions.

Limited bit depth is a common cause, but the full processing chain matters. Strong adjustments, repeated quantization, compression and display limitations can make steps more visible. A high-resolution image can still have banding because pixel count and tonal precision are different properties.

The supplied term concerns tonal banding. The same word is sometimes used for sensor stripes, lighting flicker patterns or printer defects. Those may require different diagnosis and should not automatically be treated as a shortage of gradient values.

How it works

Digital values are quantized into a finite set of representable levels. An unsigned 8-bit channel has 256 possible codes, while a 10-bit channel has 1,024. How those codes map to light depends on the encoding and color pipeline.

A gradient spanning only a small portion of the code range has fewer distinct levels available than the channel's total count. Stretching that range during editing can spread the existing steps farther apart without inventing reliable intermediate information.

Working at higher precision can reduce rounding losses during processing. Dithering adds controlled variation before or during quantization so the error appears less like coherent bands. It trades visible structure for noise-like variation; it does not make the output contain more code values than its format supports.

Three practical examples

A motion designer checks a slowly changing background gradient for visible contour-like steps after export.
A photographer examines a clear sky after strong tonal adjustments reveal bands that were not obvious in the original.
A video team compares delivery encodes to determine whether compression introduces or worsens stepping in dark, smooth regions.

Think of it this way

Think of drawing a gentle ramp using a staircase. If the steps are tall and wide enough, they become obvious instead of resembling a smooth slope. Dithering is somewhat like varying where the step transitions occur so that long, coherent boundaries become less noticeable. The analogy is limited because perception, color encoding and viewing conditions affect which digital steps are visible.

Worked example

Suppose a gradient uses only eleven stored brightness codes, from 100 through 110, across a width of 1,100 pixels. If the codes occupy roughly equal spans, each may cover about 100 pixels before the next step appears.

A later adjustment maps code 100 to 0 and code 110 to 255. Those eleven original levels remain only eleven distinct levels unless another operation estimates or introduces intermediate variation. Their separations become roughly 25.5 output-code units, making the steps much more conspicuous.

Converting this already quantized image to a higher-bit-depth container does not by itself recover the missing original gradient. Higher precision is most useful when maintained before destructive rounding or when creating a new smooth gradient from its underlying definition.

Illustrative case study

Fictional example: Eveninglight Media designs a presentation background with a subtle dark-blue gradient. It looks smooth in the editing project but shows broad bands in the exported video.

The team compares the working precision, export format and final playback path. It tests a higher-precision render and carefully applied dithering, then checks whether the delivery compression preserves the improvement. Reviewers inspect the result on the intended display rather than relying on an enlarged editor preview alone.

The source project and test exports are retained. This fictional case illustrates locating the stage where tonal information is lost, not a guarantee that one bit-depth setting fixes every visible stripe.

Common misconceptions and mistakes

Banding is not automatically caused by too few pixels. Increasing image dimensions can leave the same tonal steps intact. Saving an 8-bit image into a higher-bit-depth format also does not recreate information already lost.

Adding large amounts of noise is not a universal solution; it can distract viewers or consume compression capacity. Blurring can hide some boundaries while damaging intended detail. A smooth-looking preview on one screen does not prove the file is free of banding, and a banded preview does not prove the source data are at fault. The display and processing path should be checked.

Business use

Banding affects perceived quality in product imagery, motion graphics, video and presentation assets, particularly where the design uses broad smooth gradients. Prevention is often easier than repair after several destructive processing stages.

Maintain suitable working precision, avoid unnecessary repeated exports and inspect the final delivery format. Compare a known smooth test gradient with the actual asset to distinguish file problems from display limitations. When dithering is used, assess both visual smoothness and its behavior under the intended compression. Record the settings that work for the delivery pipeline rather than assuming a universal preset.

Check your understanding

An image containing eleven distinct gradient values is saved into a 16-bit file without other changes. Has its original smooth gradient necessarily been restored?

Answer: No. The larger container can hold more values, but the image still contains the same eleven levels unless further processing changes it.

Related terms

Gradient
Compression artifact
Tone mapping
Bit depth
quantisation
dithering
posterisation
colour precision

Learn next

Create a smooth gradient at high precision and export it using two output settings. Apply a strong contrast adjustment to copies and inspect where steps become visible. Compare versions with and without restrained dithering, then test the final files through the actual display or video-delivery path.

Gradient

Sources

Adobe, Choose the right color depth, official learning material: https://www.adobe.com/us/learn/after-effects/web/choose-the-right-color-depth Adobe Illustrator, Control dither in gradients: https://helpx.adobe.com/in/illustrator/desktop/paint-and-fill/create-and-edit-gradients/control-dither-in-gradients.html The code-range calculation and media-production case are original illustrations. Product interface details should be checked separately for the installed version.

Previous: Ball tree Next: Bandit superprocess