Why It Looks Like That
Article 11

Chroma Subsampling: What 4:4:4, 4:2:2, and 4:2:0 Really Mean

Chroma subsampling preserves full luma detail while reducing color-difference resolution. It often works well for video, but text, graphics, games, and poor reconstruction can reveal the compromise.

By · Published · Updated

Direct answer

In brief

Chroma subsampling keeps the luma-like component at full spatial resolution while reducing the width or height of the color-difference arrays. 4:4:4 keeps full chroma resolution, 4:2:2 halves horizontal chroma resolution, and 4:2:0 halves both horizontal and vertical chroma resolution. A 4:4:4 output cannot restore chroma samples missing from a 4:2:0 source, but it can move the required upsampling to a different device. Computers and games with full-resolution color generally benefit from RGB or Y′CbCr 4:4:4 when the complete chain supports it correctly.

Many video signals carry less spatial detail in their color-difference components than in their luma-like component.

That reduction is called chroma subsampling.

The most common labels are:

In current ITU definitions:

For equal bit depth and frame size, before packing or compression overhead:

The compromise often works well for photographed video because fine chromatic detail is commonly less visible than fine achromatic detail under normal viewing conditions.

It can become visible in:

The most important practical distinction is this:

Sending a 4:2:0 movie as 4:4:4 does not recreate color samples that were removed before encoding. It moves the upsampling to a different device.

A computer or game rendered with full-resolution color is different. Sending RGB or Y′CbCr 4:4:4 can preserve the source's full chroma detail if the television reproduces it correctly.

What the signal components are

A camera, renderer, or graphics system can represent an image using red, green, and blue components.

Video systems often transform those components into:

Examples include:

This separation makes chroma subsampling possible.

The system can keep the luma-like or intensity array at full spatial resolution while storing fewer color-difference samples.

A full-resolution nonlinear RGB image passes through a mathematical transform into one full-resolution Y prime luma component and two full-resolution color-difference components, Cb and Cr. A note states that Y prime is luma derived from nonlinear RGB, not physical luminance, and that the components are not literal retinal channels.
Y prime, Cb, and Cr are mathematical signal components, not literal luminance and retinal channels.

Y′ is not physical luminance

The symbol Y′ is commonly called luma.

The prime mark matters.

Y′ is derived from nonlinear R′G′B′ components. It is not the same quantity as CIE luminance Y, which is defined from linear-light tristimulus values.

Calling Y′ a brightness channel is useful for a first explanation, but it is not technically exact.

A safer description is:

Y′ carries much of the image's achromatic structure and is used as the full-resolution component in conventional Y′CbCr video.

Cb and Cr are color-difference components.

They are not direct blue and red images, and they are not literal retinal channels.

Video coding is not a copy of retinal wiring

Human vision contains multiple achromatic and chromatic mechanisms.

Video Y′CbCr is an engineering transform chosen for coding, compatibility, and efficiency.

The retina does not produce a standardized BT.709 or BT.2020 Y′CbCr signal.

The useful visual principle is narrower:

Human sensitivity to chromatic contrast and achromatic contrast differs with spatial frequency and viewing conditions.

Kathy Mullen's 1985 measurements found low-pass chromatic contrast-sensitivity functions for the tested red-green and blue-yellow gratings. Above 0.5 cycles per degree under the study's conditions, sensitivity was greater for the compared monochromatic luminance gratings than for the chromatic gratings. [9][10]

That supports reducing fine chromatic detail in many images.

It does not prove that every viewer is insensitive to every 4:2:0 artifact.

Why the compromise often works

Fine image structure is often strongly represented in the luma-like component.

Examples include:

If the color changes more slowly across those structures, reducing chroma resolution can save data with a limited visible effect.

The effectiveness depends on:

The phrase the eye sees brightness more sharply than color is a useful summary.

It is not an unconditional law.

What the ratio names mean

The colon notation comes from a historical sampling shorthand.

It is often drawn using a region four luma samples wide.

That diagram is helpful, but the modern spatial relationships are easier to understand through array dimensions.

Chroma subsampling ratios and array dimensions
Format Luma or intensity array Each chroma array
4:4:4 Full width and height Full width and height
4:2:2 Full width and height Half width, full height
4:2:0 Full width and height Half width, half height

ITU-R BT.2020-2 and BT.2100-3 define these relationships directly. [1][2]

Three side-by-side panels use the same four-by-four luma grid. The 4:4:4 panel shows four-by-four Cb and Cr arrays, 4:2:2 shows two-by-four Cb and Cr arrays, and 4:2:0 shows two-by-two Cb and Cr arrays. Labels state full chroma width and height, half width with full height, and half width with half height.
4:4:4, 4:2:2, and 4:2:0 are most clearly defined by chroma-array width and height relative to luma.

4:4:4

In 4:4:4, the two chroma arrays have the same horizontal and vertical sample counts as the luma or intensity array.

For a 3,840 × 2,160 image:

This preserves full component spatial resolution.

4:4:4 does not mean the signal is uncompressed.

A 4:4:4 image can be:

Chroma sampling and compression are separate properties.

4:2:2

In 4:2:2, each chroma array has half the horizontal samples and the same number of rows as the luma array.

For a 3,840 × 2,160 image:

The color-difference resolution is reduced horizontally.

Vertical chroma sampling remains full.

Recommendation ITU-R BT.601-7 defines a 4:2:2 family for digital studio television, illustrating the long professional use of this compromise. [4]

4:2:2 remains useful where a workflow wants lower data than 4:4:4 but stronger chroma detail than 4:2:0.

4:2:0

In 4:2:0, each chroma array has half the horizontal samples and half the vertical samples of the luma array.

For a 3,840 × 2,160 image:

Each chroma array contains one-quarter as many samples as the luma array.

This is the source of the phrase quarter-resolution chroma.

It refers to sample count, not to one-quarter of the image width or one-quarter of the visible color gamut.

The second row is not displayed without color

A common diagram says that 4:2:0 samples chroma on one row and takes no chroma samples on the next.

That description is easy to misread.

The bitstream stores lower-resolution chroma arrays.

The decoder or display reconstructs chroma at the working pixel positions.

Every displayed pixel still receives color information after upsampling.

The reconstructed value may be calculated from:

The lower-resolution array limits the original spatial detail.

It does not produce a visibly colorless alternating row.

A signal path begins with a four-by-four luma grid and two-by-two stored Cb and Cr grids, passes through a chroma upsampling filter, produces full working Cb and Cr grids, and then converts to final full-color RGB pixels. The diagram states that every displayed pixel receives reconstructed color information and that the lower-resolution source limits the original chroma detail.
A 4:2:0 image is displayed after chroma reconstruction, not as alternating color and colorless rows.

Chroma sample siting matters

A lower-resolution chroma sample represents a position relative to the luma grid.

That position is called chroma sample siting or chroma sample location.

4:2:0 has more than one possible siting convention.

Recommendation ITU-T H.274 V4 defines several location types using horizontal and vertical offsets. It also gives specific guidance for decoded content intended for BT.2020 or BT.2100 interpretation. [3]

If a conversion assumes the wrong siting, chroma can shift by part of a pixel.

Possible visible effects include:

Correct upsampling needs the right geometry as well as the right sample values.

Three conceptual four-by-four luma grids show different possible four-to-two-to-zero chroma sample locations: co-sited with the top-left luma position, horizontally aligned with the left luma column and vertically centered between rows, and centered within each two-by-two luma region. The figure states that current coded-video standards can signal multiple chroma sample locations and that correct reconstruction needs the signaled geometry.
The 4:2:0 ratio fixes array dimensions, while sample siting separately defines the chroma geometry.

How many raw samples are saved?

Consider a 2 × 2 block of four luma positions.

4:4:4

4:2:2

8 / 12 = 2/3

4:2:0

6 / 12 = 1/2

These comparisons assume:

They are raw sample-count comparisons.

A compressed 4:2:0 file is not guaranteed to be exactly half the size or bitrate of a comparable 4:4:4 file.

For one two-by-two luma block, three cards show 4:4:4 with four luma, four Cb, and four Cr samples for a total of twelve; 4:2:2 with four luma, two Cb, and two Cr samples for a total of eight; and 4:2:0 with four luma, one Cb, and one Cr sample for a total of six. The figure states that 4:2:2 uses two-thirds and 4:2:0 uses one-half of the 4:4:4 raw three-component sample count under equal bit depth and no packing or compression overhead.
4:2:2 uses two-thirds and 4:2:0 uses one-half of the 4:4:4 raw component-sample count under stated conditions.

Sampling and compression are different

Chroma subsampling is a spatial sampling choice.

Compression is a coding process that removes statistical or perceptual redundancy and represents the remaining data more efficiently.

A signal can be:

The two decisions interact, but they should not be treated as the same thing.

Where 4:2:0 is common

Many consumer delivery profiles use 4:2:0.

The CTA HDR10 Media Profile specifies 4:2:0 for compressed video sources, along with ten-bit depth and BT.2020 primaries. [5]

4:2:0 is attractive for distribution because it:

This does not mean every consumer signal or every codec uses only 4:2:0.

Codec name does not prove chroma format

A codec family can support multiple chroma formats through different profiles and constraints.

Recommendation ITU-T H.264 includes higher-chroma profiles in addition to common 4:2:0 profiles. [11]

HEVC and other modern codec families also have profile-dependent capabilities.

A format claim should identify:

The name H.264, HEVC, or AV1 by itself is not proof of 4:2:0.

Where 4:2:2 and 4:4:4 are common

Higher chroma resolution may be chosen for:

The choice depends on the workflow.

4:4:4 is not automatically required at every stage, and 4:2:0 is not automatically acceptable at every stage.

Why photographed video often tolerates 4:2:0

Photographed scenes contain:

These features can mask small chroma-resolution losses.

Many edges are also defined strongly in Y′.

At a normal viewing distance, the reconstructed chroma can appear aligned well enough that the viewer notices the full-resolution structure rather than the lower chroma grid.

This is a tendency, not a guarantee.

A saturated object against an isoluminant background can expose chroma loss even in a photographed scene.

When text reveals chroma subsampling

Small text is demanding because it can contain:

Black text on white is mostly an achromatic test and may remain sharp even with lower chroma resolution.

Colored text on a contrasting colored or gray background can reveal the chroma reduction more clearly.

Possible symptoms include:

A proper test pattern includes combinations that stress both horizontal and vertical chroma resolution.

Two conceptual panels compare a photographed scene with texture, mixed edges, shading, and motion cues against chroma-sensitive graphics containing small colored text, thin saturated lines, a game HUD, and hard flat-color boundaries. A qualification states that visibility depends on chroma contrast, viewing distance, sample siting, compression, scaling, and reconstruction.
Fine colored text and hard saturated edges can reveal chroma loss more readily than many textured scenes.

Why computer desktops benefit from full chroma

A desktop operating system can render every pixel with independent R, G, and B values.

Small text and interface elements are designed for a full-resolution pixel grid.

If the source converts that desktop to Y′CbCr 4:2:0, fine chromatic information is reduced before the television receives it.

For a television used as a monitor, a good goal is:

The television must also preserve that full chroma through its input mode, scaling, and processing.

Two computer-to-television paths compare a full-resolution RGB desktop sent as RGB or Y prime CbCr four-to-four-to-four into a television mode that preserves full chroma, with crisp colored text, against a four-to-two-to-zero path that reduces fine colored text and line detail before the television receives it. A note says the television must also preserve one-to-one scaling and full chroma at the selected timing.
Full chroma can preserve real desktop and game-interface detail when the complete display path supports it.

Games contain both video-like and desktop-like material

A game scene may tolerate chroma reduction in:

The same game may contain chroma-sensitive elements:

Full chroma can therefore matter even when the main moving scene looks similar.

High refresh rate and HDR may create a bandwidth tradeoff. The best format depends on the complete source, cable, receiver, and display capabilities.

Animation is not one category

Animation can contain:

It can also contain:

Some animation tolerates 4:2:0 well.

Some exposes it quickly.

The art and encoding matter more than the label animation.

Chroma upsampling is unavoidable for a 4:2:0 source

A modern flat-panel display needs color values at its output pixels.

A 4:2:0 source has lower-resolution chroma arrays.

Someone must upsample them.

The upsampling can occur in:

The chosen HDMI output format helps determine where the conversion happens.

Why outputting 4:4:4 does not restore a 4:2:0 master

Suppose a movie is encoded as 4:2:0.

The source device decodes the lower-resolution chroma arrays.

If it outputs 4:4:4, it must create the additional chroma samples through interpolation.

Those samples are estimates.

The device cannot know the exact full-resolution chroma that existed before subsampling.

A 4:4:4 output can still look better when the source device's chroma upsampling is better than the television's.

It can look worse when:

The output label alone does not establish quality.

A four-stage chain starts with an original full-chroma image, downsamples it into a four-to-two-to-zero master, upsamples that master in a player to four-to-four-to-four with estimated samples marked, and sends the full-resolution output to a television. A note states that the output choice can move the reconstruction between devices, but cannot recreate unknown original chroma samples.
A 4:4:4 output can move chroma upsampling to another device, but cannot recreate unknown source samples.

Chroma upsampling quality matters

A basic filter may average nearby samples.

A more advanced filter may consider:

Poor processing can create:

Recommendation ITU-T H-Series Supplement 15 addresses conversion and coding practices for HDR and wide-color-gamut Y′CbCr 4:2:0 PQ video, reflecting the care required in these transforms. [12]

RGB and Y′CbCr answer different questions

RGB describes three primary-color components.

Y′CbCr describes one luma-like component and two color-difference components.

A source can output:

RGB and Y′CbCr 4:4:4 both retain full spatial resolution in their three component arrays.

They are not numerically identical.

Conversion requires the correct:

A correct conversion can be accurate to the intended representation.

An incorrect conversion can shift levels or colors.

RGB is not automatically superior

For a computer desktop, RGB is often the most direct output.

For video content, the source may already be stored as Y′CbCr 4:2:0.

The path to the panel may involve one or more conversions regardless of the HDMI output choice.

The best output depends on:

Use Auto as a starting point when the devices negotiate correctly.

Use a test pattern when the visible result matters.

Full chroma is not full range

A source menu may combine terms that sound similar.

They are separate.

Chroma sampling

Signal range

Possible combinations include:

Both source and display must interpret the range the same way.

A mismatch can produce:

That problem is not chroma subsampling.

A two-axis matrix places chroma sampling on the horizontal axis with four-to-four-to-four, four-to-two-to-two, and four-to-two-to-zero, and signal range on the vertical axis with full and narrow. Example cards include RGB full, RGB limited, Y prime CbCr four-to-four-to-four narrow, and Y prime CbCr four-to-two-to-two narrow. A warning states that a range mismatch can crush or raise blacks even when chroma is full resolution.
Chroma sampling and digital range are independent, and both must be interpreted correctly.

PC Mode is manufacturer-specific

A television may require an input to be labeled PC to preserve full chroma.

Another may preserve 4:4:4 in a normal picture mode.

Another may support full chroma only at selected timings or refresh rates.

PC Mode may change:

The name does not prove the measured result.

Use the exact manual and a full-chroma test pattern.

Game Mode is not the same as PC Mode

Game Mode is designed primarily to reduce latency.

It may preserve 4:4:4.

It may not.

A television can have:

Do not use one model's behavior as a rule for another.

Enhanced input modes affect several capabilities

Manufacturers use names such as:

These modes can enable higher-bandwidth signal combinations.

They may affect:

The exact effect depends on the model, input, and firmware.

HDMI bandwidth creates tradeoffs

An uncompressed link must carry all selected component samples.

Bandwidth rises with:

Reducing chroma lowers the required data rate.

This can allow a system to preserve:

within the same link limit.

An 18 Gbit/s TMDS-class link cannot carry every 4K60 full-chroma deep-color combination.

For example, uncompressed 4K60 RGB or Y′CbCr 4:4:4 at ten bits exceeds that link's 18 Gbit/s maximum.

Common alternatives include:

The correct tradeoff depends on the content.

A central HDMI link budget is surrounded by resolution, frame rate, bit depth, and chroma sampling. Three boundary cards identify Premium High Speed at 18 gigabits per second, Ultra High Speed up to 48 gigabits per second, and Ultra96 up to 96 gigabits per second. The figure states that product support depends on named features and port capabilities, not a version number alone, and that 4K60 four-to-four-to-four ten-bit exceeds an 18 gigabit per second TMDS-class link.
Resolution, frame rate, bit depth, and chroma sampling share the available HDMI link bandwidth.

Do not shop by HDMI version number alone

HDMI Licensing Administrator states that specification version numbers are not feature sets because many features are optional. [7]

A product called HDMI 2.1 does not necessarily support:

Check the named features and actual format support.

The same principle applies to later versions.

Official cable categories

HDMI Licensing Administrator identifies:

Use the cable category required by the actual source and display format.

HDMI version numbers should not be used as cable product names.

A certified cable does not add capabilities that the devices lack.

Practical setup for a computer

A useful starting point is:

  1. Use the television input documented for PC use or high bandwidth.
  2. Enable the required enhanced input mode.
  3. Select RGB or Y′CbCr 4:4:4.
  4. Match full or limited range at both ends.
  5. Use the native display resolution.
  6. Disable overscan.
  7. Confirm the refresh rate.
  8. Display a full-chroma text pattern.
  9. Check SDR and HDR separately.

Do not copy another television's menu values.

Practical setup for a game console

A useful starting point is:

  1. Enable the television's documented game or low-latency mode.
  2. Enable the required enhanced input format.
  3. Let the console detect supported formats.
  4. Preserve ten-bit HDR where required.
  5. Use full chroma when bandwidth and the full chain support it.
  6. Accept 4:2:2 or 4:2:0 when necessary to preserve a more important feature such as HDR bit depth or high refresh rate.
  7. Match range at both ends.
  8. Test UI text and real gameplay.
  9. Verify each refresh-rate mode separately.

The highest chroma number is not automatically the best complete configuration.

Practical setup for a movie player

A movie encoded as 4:2:0 must be upsampled before display.

Possible output choices can all be valid:

Choose the path that:

Do not assume that 4:4:4 output means the disc or stream contained 4:4:4 source chroma.

Practical setup for a streaming box or cable receiver

Start with Auto when it correctly matches the source and display.

Then verify:

Some devices use one fixed output format for menus and video.

That can force conversion even when the program itself was encoded differently.

The best choice is device-specific.

A four-row setup matrix compares computer, game console, movie player, and streaming box. Columns show full-chroma priority, frame-rate priority, HDR bit-depth priority, range matching, and whether a test is needed. The matrix gives high full-chroma priority for computers, balanced priorities for games, reconstruction-path testing for movie players, and device-specific verification for streaming boxes. A footer states that there is no universal best RGB or Y prime CbCr output for every source.
Computers, games, movie players, and streaming boxes place different priorities on chroma, rate, and HDR depth.

How to test 4:4:4 reproduction

A useful full-chroma test pattern includes:

The test should be:

A web image can be altered by:

A test image copied from a social platform is weak evidence.

What a failed chroma test can mean

Soft or colored text can result from:

Identify the signal before blaming the panel.

What calibration can verify

Controlled testing can evaluate:

Record:

What calibration cannot restore

Calibration cannot recreate chroma samples that were discarded before encoding.

It cannot:

Calibration can reduce measurable display error and help identify which path reproduces the source most faithfully within the system's capabilities.

Where this leaves us

Chroma subsampling reduces spatial detail in color-difference components.

4:4:4 keeps full chroma array dimensions.

4:2:2 halves horizontal chroma resolution.

4:2:0 halves both horizontal and vertical chroma resolution, leaving each chroma array with one-quarter of the luma sample count.

For equal bit depth and frame size, 4:2:0 uses half the raw three-component sample count of 4:4:4.

Y′ is luma, not physical luminance.

Y′CbCr is an engineering representation, not a direct model of retinal wiring.

4:2:0 does not mean every second row is displayed without color.

The decoder reconstructs chroma using the stored samples, their positions, and an interpolation process.

Many consumer delivery profiles use 4:2:0.

Not every codec, interface, or production workflow does.

Photographed video often tolerates the reduction.

Small colored text, graphics, games, and hard-edged animation can reveal it.

A 4:4:4 output cannot restore chroma samples missing from a 4:2:0 master.

It can change which device performs the reconstruction.

RGB is not automatically superior to correctly processed Y′CbCr 4:4:4.

Full chroma and full range are different settings.

PC Mode, Game Mode, and enhanced HDMI modes vary by television.

HDMI version numbers do not guarantee one feature set.

The useful question is not simply:

Is 4:4:4 better?

It is:

Did the source contain full chroma, where was chroma reduced or reconstructed, does the content need fine chromatic detail, and which complete signal path reproduces it most accurately?

Sources and standards

Primary video-sampling standards, HDMI documentation, color-vision research, codec guidance, and HDR/WCG conversion guidance used in this article. Last verified July 31, 2026.

  1. International Telecommunication Union Radiocommunication Sector. Parameter Values for Ultra-High Definition Television Systems for Production and International Programme Exchange. Recommendation ITU-R BT.2020-2, October 2015.
    https://www.itu.int/dms_pubrec/itu-r/rec/bt/R-REC-BT.2020-2-201510-I%21%21PDF-E.pdf

    ↩↩↩

  2. International Telecommunication Union Radiocommunication Sector. Image Parameter Values for High Dynamic Range Television for Use in Production and International Programme Exchange. Recommendation ITU-R BT.2100-3, February 2025.
    https://www.itu.int/dms_pubrec/itu-r/rec/bt/R-REC-BT.2100-3-202502-I%21%21PDF-E.pdf

    ↩↩↩

  3. International Telecommunication Union Telecommunication Standardization Sector. Versatile Supplemental Enhancement Information Messages for Coded Video Bitstreams. Recommendation ITU-T H.274 V4, January 2026.
    https://www.itu.int/epublications/publication/itu-t-h-274-v4-2026-01-versatile-supplemental-enhancement-information-messages-for-coded-video-bitstreams

    ↩↩

  4. International Telecommunication Union Radiocommunication Sector. Studio Encoding Parameters of Digital Television for Standard 4:3 and Wide-Screen 16:9 Aspect Ratios. Recommendation ITU-R BT.601-7, March 2011.
    https://www.itu.int/dms_pubrec/itu-r/rec/bt/r-rec-bt.601-7-201103-i%21%21pdf-e.pdf

    ↩

  5. Consumer Technology Association Video Division. 4K Ultra High-Definition TV and High Dynamic Range Compatible Display Consumer Definitions.
    https://cdn.cta.tech/cta/media/media/membership/pdfs/video-technology-consumer-definitions.pdf

    ↩

  6. HDMI Licensing Administrator, Inc. HDMI Cables: Different Cable Types.
    https://www.hdmi.org/resource/cables

    ↩

  7. HDMI Licensing Administrator, Inc. What Does It Mean When Gear Is Marketed as an HDMI 2.1 Product? June 16, 2021.
    https://www.hdmi.org/blog/detail/139

    ↩

  8. HDMI Licensing Administrator, Inc. HDMI 2.2 Specification Technology Overview.
    https://www.hdmi.org/spec/hdmi2

    ↩

  9. Mullen, Kathy T. The Contrast Sensitivity of Human Colour Vision to Red-Green and Blue-Yellow Chromatic Gratings. The Journal of Physiology 359, pages 381-400, February 1985. DOI 10.1113/jphysiol.1985.sp015591.
    https://doi.org/10.1113/jphysiol.1985.sp015591

    ↩

  10. International Telecommunication Union Radiocommunication Sector. The Present State of Ultra-High Definition Television. Report ITU-R BT.2246-9, September 2025.
    https://www.itu.int/dms_pub/itu-r/opb/rep/R-REP-BT.2246-9-2025-PDF-E.pdf

    ↩

  11. International Telecommunication Union Telecommunication Standardization Sector. Advanced Video Coding for Generic Audiovisual Services. Recommendation ITU-T H.264.
    https://www.itu.int/rec/t-rec-h.264

    ↩

  12. International Telecommunication Union Telecommunication Standardization Sector. Conversion and Coding Practices for HDR/WCG Y′CbCr 4:2:0 Video with PQ Transfer Characteristics. Supplement 15 to the H-series, January 2017.
    https://www.itu.int/dms_pubrec/itu-t/rec/h/T-REC-H.Sup15-201701-I%21%21SUM-HTM-E.htm

    ↩
Related reading: TV Test Patterns: What They Show and How to Use Them Use chroma-resolution and other patterns to check what the full playback path is preserving. Next step: HDR Formats Explained Move from chroma sampling into HDR10, HDR10+, Dolby Vision, HLG, metadata, compatibility, and format selection.