Resize a JPG Image

When a resize genuinely improves a JPG

The reason resizing a JPG is often worthwhile is that a camera saves at full resolution, and most places an image is used display only a fraction of that. A 12 megapixel photo at 6,000 by 4,000 displayed at 1,200 pixels wide is carrying five times the pixels anyone will see, and a file several times larger than it needs to be. Reducing the dimensions to roughly twice the display width gives a sharp result on a high-DPI screen with a large saving, and nothing visible is lost. That is a genuine improvement rather than a compromise.

NoteTwice the display width covers high-DPI screens. Beyond that, extra pixels are rarely visible.

Reducing versus enlarging are completely different

Resizing down discards pixels, and the result is sharp because what remains was always there. Resizing up has to invent pixels, and the result is interpolated, which is a guess rather than recovered detail. Sharpening after an enlargement can make the edges look more defined, but it does not add information and it creates haloes. The practical rule is that a resize below the native size is safe and useful, and a resize above it should only be done when the exact dimensions are required and there is no alternative.

• Reducing: discards pixels, result stays sharp. Always safe.
• Enlarging: invents pixels, result is softer. Avoid where possible.
• A warning appears when a source is enlarged to reach a target.
• Sharpening an enlargement looks crisper but is not sharper.

NoteA reduction is a genuine improvement. An enlargement is a compromise you may be forced into.

Choosing a mode, and why the default matters

A non square image has to be mapped onto a new shape, and the three modes trade differently. Crop to fill covers the frame and trims the overflow, which suits a landscape photo going to a banner. Fit keeps the whole image and pads the remainder, which suits a portrait photo going to a square, since a fill would cut the head off. Stretch forces the dimensions and distorts, which almost never looks right for a photograph. The default is usually fine for a simple reduction, and worth changing when the shape is changing too.

• Fill: covers the frame, trims the overflow. Good for a photo into a banner.
• Fit: keeps everything, pads with a solid colour. Good for a portrait into a square.
• Stretch: forces the shape and distorts. Avoid for photographs.
• The default is fine for a same shape reduction, not for a shape change.

JPG quality is a second, separate decision

Resizing changes the pixels and does not by itself change the quality, but saving the result is a lossy step. At a balanced setting the double loss of reducing then re-encoding is not usually visible, and the file is small. At a high setting the result is cleaner when zoomed, at the cost of a larger file. A useful check is to compare the resized output against the original at 100 percent at the new size, rather than magnified, because a reduction viewed zoomed will look softer than it does in use.

• Resizing and re-encoding are two separate lossy steps.
• Balanced quality is usually indistinguishable at display size.
• Compare at 100 percent at the new size, not magnified.
• Keep the original. A different size later needs it.

NoteCompare at the new size and 100 percent. Zoomed comparison is misleading.