Does Compressing an Image Reduce Quality? Six Myths Checked

2026-10-11

Does compressing an image reduce quality? Sometimes it does and sometimes it does not, and the difference comes down to which kind ran. Lossy compression discards high frequency detail for good. Lossless compression only rewrites how the same bytes are stored, so every pixel comes back identical. What wrecks a photo is rarely one operation in isolation. It is the same file being re-encoded by three platforms in a row until blocking artifacts stack up.

Lossy and lossless discard different things

• Lossy compression drops detail the eye struggles to notice, including the colour information that chroma subsampling halves or quarters
• Lossless compression changes nothing but the encoding, so PNG, ZIP and lossless WebP restore every pixel exactly
• The quality slider on a JPEG export controls how aggressive the quantisation table is, and even 100 stays lossy
• Lossy WebP and AVIF discard data too, they just hide the damage better at the same file size
• HEIC borrows video compression machinery and pays for its small files the same way

One question settles it: can the pixels be restored exactly? If yes, the route was lossless. If no, data left the building. Most web images take the lossy route, because that is what halves the file.

What one pass actually costs

Take a JPEG straight out of a camera down to 60 or 75 percent quality, zoom to actual pixels and inspect a flat area. Faint banding shows up in a sky or a painted wall. Viewed at the size the image actually occupies on a page, the same difference is invisible.

• 90 to 100: barely smaller, and rarely worth applying to a JPEG
• 75 to 85: the range where most photographs resist close inspection
• Below 60: ringing and blocking arrive around leaves, hair and text edges
• Large smooth gradients expose artifacts first, so check sky and walls
• Fine texture is where encoders cut hardest: insect wings, fabric, fur
• Screenshots and anything carrying text belong in PNG or lossless WebP

The complaint that compression ruined the picture is legitimate. Its triggers are just narrower than most people assume: either the quality setting went too low, or the image belongs in a lossless format.

Six claims, checked one at a time

• Every kind of compression costs quality. Wrong. Lossless modes restore every pixel, so a PNG can be recompressed indefinitely without accumulating damage.
• Quality 100 is the original file. Wrong. It is the lightest setting JPEG offers, re-encoding has already happened, and the result is often larger than the source.
• Compress ten times, lose ten times as much. Not quite. Damage accumulates but the increments shrink, so passes two and three hurt far more than pass eight.
• Resizing is compression. No. Changing the pixel count removes detail permanently, which is a different operation from re-encoding, and /blog/compressed-vs-resized-image separates the two.
• PNG never loses quality. Correct. Making a PNG smaller means fewer colours or a stronger encoder, not discarded data.
• WebP always beats JPEG. Usually, but not automatically. WebP holds more detail at equal size only when both sides were compressed once; converting an existing JPEG and compressing again leaves a blurrier file.

The pattern across all six: damage traces back to specific choices, not to the word compression. Pick the format, limit the passes, set a sensible quality number, and whatever is lost stays below the threshold anyone notices.

Repeated compression is what kills photos

• One platform compresses on upload, another recompresses on republish, a chat app does it again on send, and by the third round the edges fringe
• Each pass treats the blocking left by the previous pass as genuine detail, then smooths over that
• Social apps usually re-encode to a fixed low setting, and choosing original quality on upload is the way around it
• Export each derivative once from the master instead of re-saving an already compressed version
• Long images and animated frames degrade faster, since differences between frames get flattened into each other

Compress a given image once. Every size, format and upload should derive from that single export rather than from versions of versions.

The mechanics are laid out in /blog/lossy-vs-lossless-compression, where the chroma information gets cut in /blog/chroma-subsampling-explained, resize versus compression ordering in /blog/resize-and-compress-image, and how resolution maps to file size in /blog/image-resolution-vs-file-size.

FAQ