JPG Chroma Subsampling: 4:4:4 vs 4:2:2 vs 4:2:0

Why Colored Edges Smear in JPEGs

If you have saved a screenshot of a dark UI or a scanned document as a JPEG and noticed red, green, or blue fringing along black text, you have seen chroma subsampling. The same mechanism makes thin colored lines on a white background appear to bleed, and it turns small red text into an unreadable blur. This artifact is independent of the quality setting. A JPEG saved at 100% quality can still exhibit the smear because the damage occurs before compression, when the encoder decides how many color samples to store.

JPEG encoders convert RGB pixels into YCbCr, where Y carries luma (brightness) and Cb/Cr carry chroma (color-difference signals). The human visual system detects luminance changes at higher spatial frequencies than color changes. The JPEG standard (ITU-T.81, Annex A) permits chroma channels to be stored at a lower resolution than luma. That resolution reduction is chroma subsampling. The pattern chosen determines whether every pixel retains its original color or whether groups of pixels share a single color value. That decision is fixed at encode time and cannot be reversed by a decoder.

How Subsampling Differs from the Quality Parameter

JPEG quality—expressed as a percentage, a Q-factor, or a quantization table index—controls the quantization step that discards high-frequency DCT coefficients. It reduces detail and introduces blocking artifacts, but it applies to luma and chroma alike. Chroma subsampling determines how many color samples exist before quantization begins.

A 4:4:4 file stores full chroma for every pixel, then compresses that data according to the quality setting. A 4:2:0 file discards three-quarters of the chroma samples first, then compresses the remainder. This is why a 4:2:0 JPEG at quality 98 can look acceptable for a photograph yet destroy a black-on-white line chart. The quality setting preserved the luma edges, but the chroma channels were decimated, so colored edges are approximated across 2×2 pixel blocks. No increase in quality can recover the discarded samples. The only remedy is to re-encode from an uncompressed source with 4:4:4 subsampling.

The Three Subsampling Patterns

4:4:4 — Full Chroma Resolution

Every pixel carries its own Cb and Cr pair. Chroma resolution equals luma resolution. This is the only mode that preserves sharp colored edges, anti-aliased fonts, and UI elements with colored borders.

When to use 4:4:4:

• Screenshots of web pages, code editors, or design tools where text and UI colors must remain crisp
• Line art, diagrams, and charts with colored labels
• Images destined for further editing or re-compression
• Legal or evidentiary images where pixel-level color accuracy is required

Cost: File size increases roughly 1.5× to 2× compared with 4:2:0 at the same quality setting because three times as much chroma data is stored. Many web browsers and social platforms re-encode uploads to 4:2:0, so 4:4:4 is most useful for local archiving, printing, or direct delivery to a recipient who will not re-compress.

4:2:2 — Half Horizontal Chroma

Each 2×1 block of pixels shares one Cb/Cr pair. Chroma resolution is halved horizontally but kept full vertically. This pattern originated in broadcast video and digital cinema. For still images, it occupies a narrow middle ground.

When to use 4:2:2:

• Images containing strong vertical color transitions, such as colored vertical stripes
• Scanning workflows where horizontal resolution is the limiting factor
• Situations that require some file-size reduction but cannot tolerate the worst 4:2:0 artifacts on horizontal color edges

Cost: File size is about 1.33× larger than 4:2:0. The visual improvement over 4:2:0 is marginal for most photographs. The trade-off is rarely justified unless you are specifically avoiding horizontal color smearing.

4:2:0 — Quarter-Resolution Chroma

Every 2×2 block of pixels shares one Cb/Cr pair. Chroma resolution is halved in both directions. This is the default for nearly all consumer cameras, web platforms, and social media services. It reduces file size significantly with minimal visible impact on natural photographs because real-world color gradients change slowly.

When to use 4:2:0:

• Photographs of landscapes, portraits, products, or still life where fine color edges are not critical
• Web delivery where bandwidth and load time outweigh pixel-perfect color
• Bulk archiving of family photos where storage space is a concern
• Any image that will be re-encoded by a platform (Instagram, Facebook, Twitter) that applies 4:2:0 regardless of input

When to avoid 4:2:0: Any image with synthetic content—text, UI elements, logos, line art, screenshots, or scanned documents with colored highlights.

Diagnosing Subsampling in an Existing File

You cannot determine subsampling from the file extension or the quality number. You must inspect the chroma planes.

1. Open the JPEG in an image editor that can display individual channels.
2. Examine the Cb and Cr channels (sometimes labeled U and V).
3. If those channels appear at half or quarter resolution—blocky, pixelated, or visibly smaller than the Y channel—subsampling is present.
4. For a precise read, use a tool that reports JPEG sampling factors. ExifTool shows ChromaSubsampling: YCbCr4:2:0 (2 2) or similar. ImageMagick’s identify -verbose reports Sampling factors directly.

What Conversion Cannot Recover

Once a JPEG has been saved with 4:2:0 subsampling, the discarded color information is absent from the file. No re-compression, upsampling, or AI-based sharpening can reconstruct the original per-pixel color values. Decoders interpolate chroma back to full resolution when the image is opened, but the interpolation is a guess. The red fringe around black text persists because the chroma was averaged across the 2×2 block that contained both the black pixel and the white background.

The only recovery path is to return to the original source—the RAW file, the screenshot capture, the SVG, or the layered document—and export again at 4:4:4.

How to Export with 4:4:4 Subsampling

Most consumer tools default to 4:2:0. You must explicitly override the subsampling setting.

• Adobe Photoshop: In Save for Web (Legacy), select JPEG, set Quality, click the small gear icon, and disable “Optimized” (this preserves 4:4:4 in many versions). In the standard Save As dialog, uncheck “Progressive”; subsampling may still default to 4:2:0 unless you use Save a Copy with “Maximum compatibility.”
• ImageMagick: convert input.png -sampling-factor 4:4:4 -quality 92 output.jpg
• libjpeg / mozjpeg (command line): cjpeg -sample 1x1 -quality 92 input.ppm > output.jpg (1×1 means no subsampling; 2×2 means 4:2:0)
• Adobe Lightroom Classic: Export → File Settings → set Quality. Under the “Other” preset, set Chroma Subsampling to “4:4:4 (Best Quality)” in the advanced options. The standard presets do not expose this control.
• macOS Preview: Provides no user control and always exports 4:2:0. Do not use it for text-heavy JPEGs.

When 4:4:4 Is Non-Negotiable

If you are preparing images for a print workflow that involves offset printing or a professional photo lab, and the image contains synthetic elements such as text, logos, or colored lines, 4:4:4 is the only safe choice. The raster image processor (RIP) at the print shop applies its own color management, and any chroma averaging from 4:2:0 will be magnified at the printer’s resolution, producing visible halos around every letter and graphic element. The JPEG standard (ITU-T.81, Annex A) defines the encoding process but leaves decoding to the application; many print RIPs decode 4:2:0 with bilinear interpolation that exacerbates the artifacts. Send a 4:4:4 JPEG or, preferably, a TIFF with no subsampling at all.

Technical basis for JPG Chroma Subsampling

Structural claims in this guide are grounded in the primary JPEG technical reference.