Image quality settings explained: what the number really does

What the quality number controls

JPEG compression works by throwing away details we weren't likely to see in the first place. The quality setting tells the encoder how much it may throw away. High numbers keep more detail and give bigger files; low numbers discard more and give smaller ones.

The scale is usually 0 to 100, but it is not a percentage of anything. A quality of 80 doesn't mean 80 % of the detail is kept or that the file is 80 % of the original size. Different programs don't even agree on what a number means: not every tool interprets a value of 75 the same way. So a setting that worked in one app is only a rough starting point in another.

Why the same quality gives different sizes

The size of a compressed image depends on its content as much as on the setting. A photo of a clear sky has large smooth areas that take very little data to describe. A photo of gravel, leaves or a crowd is full of fine detail, and every bit of it costs data. At the same quality setting, the second file comes out noticeably larger than the first.

For the same reason, the visible effect of a given setting varies from picture to picture. This is why no single number is right for everything, and why a fixed figure such as 'quality 70 halves the size' can't be promised.

What compression artefacts look like

When the quality goes too low, the missing detail becomes obvious. JPEG works on small blocks of pixels, so the first sign is often faint square seams, especially in skies and other smooth areas. Other common signs:

• Smudged fine texture, such as hair or grass turning into flat patches.
• Blocky steps where one colour changes into another.
• Colour fringes that bleed slightly across sharp edges.

Colour detail: chroma subsampling in plain words

Our eyes are much better at seeing fine changes in brightness than fine changes in colour. JPEG takes advantage of this. It separates the brightness of each pixel from its colour, keeps the brightness at full resolution, and often stores the colour at a lower resolution. A common scheme keeps one colour value for each block of 2 × 2 pixels, while every pixel keeps its own brightness. This is called chroma subsampling.

On ordinary photos you rarely notice it. It shows on very fine coloured detail: thin red lines or small coloured text can turn duller or blurred at the edges. That is another reason screenshots and graphics with coloured text are better saved as PNG.

Re-saving a JPG loses a little more each time

Each time you open a JPG, edit it and save it again as a JPG, the compression runs again on a picture that has already lost detail. What was thrown away doesn't come back, and new small errors are added. Over several rounds this generation loss becomes visible, a bit like photocopying a photocopy. Resizing or rotating between saves makes it worse.

The practical rule: compress once, at the end, from the best original you have. Keep the original or a lossless master for editing, and make compressed copies from it rather than compressing an already compressed copy.

Lossless optimisation for PNG

PNG is lossless, so there is no quality dial to lower without changing the picture. Even so, different programs encode PNG with very different efficiency. An optimiser can rewrite the same pixels more compactly, giving a smaller file with an identical image.

The image compressor does this with PNG files, and uses lossy encoders for JPG, WebP and AVIF. If the result isn't smaller than your file, it keeps the original rather than handing back a bigger one. For a photo saved as PNG, switching Save as to JPG or WebP usually saves far more: the same photo can be a few hundred kilobytes as a JPEG but several megabytes as a PNG.