Sharpen Image

Sharpening cannot unblur a photo

Worth settling first, because most of what you will read about this promises otherwise.

Sharpening does not recover detail. It finds edges — places where light meets dark — and exaggerates the difference, making the dark side darker and the light side lighter. Your eye reads that increased contrast as crispness. Nothing new has been added; the existing information has been pushed further apart.

Which tells you exactly when it works and when it does not.

It works on an image that is slightly soft but fundamentally in focus: a photo that was downscaled (shrinking an image always softens it), a scan, a screenshot, a shot taken through glass, a picture that came out of a phone's processing looking a little flat.

It does not work on a photo that is genuinely out of focus, or blurred by camera shake or subject movement. In those images the detail was never recorded — the light from one point landed across many pixels. Sharpening such a photo raises the contrast of the blur itself, which looks harsher and no clearer. There is nothing to sharpen.

If that is your situation, the honest answer is to retake the shot if you can. Tools that claim to unblur with AI are guessing at plausible detail rather than recovering real detail, and on a face or any text that guess is often visibly wrong.

"Edge Enhance" is the gentlest option, not the strongest

The name is misleading and it matters, so here is what the three methods actually do.

Standard compares each pixel with its four neighbours — above, below, left, right — and pushes it away from their average. This is the classic sharpening filter and the right starting point.

Strong uses all eight surrounding pixels, diagonals included. At the same intensity setting it is roughly twice as aggressive, and because it includes the diagonals it bites harder on diagonal edges and fine texture. Use it on an image that needs real help.

Edge Enhance is the same filter as Standard at half the strength. So despite the name, it is the subtlest of the three. That makes it the right choice for the situations where sharpening usually goes wrong — portraits, skin, anything with smooth gradients, and any image that is only a little soft.

Practically: start with Edge Enhance or Standard at the default intensity of 3. Move to Strong only if the result is clearly not enough.

Start low, because over-sharpening is obvious

The most common mistake, and it looks worse than the original problem.

Push the intensity too far and three things appear. Halos — bright outlines along high-contrast edges, most visible where something dark meets sky. Crunchy texture in areas that should be smooth: skin becomes mottled, a clear sky develops grain. And jagged diagonals, where a smooth angled line breaks into visible steps.

Those artefacts are cumulative and permanent once saved. Work up from 3 rather than down from 10, and judge the result at full size — sharpening looks fine on a thumbnail and awful at 100%.

Two things sharpening makes worse

Both follow from the same mechanism, and neither is usually mentioned.

Noise. Sensor noise is fine, high-contrast, pixel-level variation — which is exactly what a sharpening filter is designed to amplify. A photo shot at high ISO in low light will come back noticeably noisier. If an image is both noisy and soft, reduce the noise first; sharpening first locks the noise in.

JPEG artefacts. A heavily compressed photo contains 8×8 block boundaries and faint ringing around edges. Sharpening raises the contrast of those too, so the compression becomes visible where it was not before. On a photo that has been saved and re-saved several times, sharpening often makes it look worse overall even where the subject looks crisper.

Sharpen last

The ordering rule that makes the biggest difference to the result.

Sharpening is tied to pixel size, so anything that changes the pixels afterwards undoes or distorts it. Resize first, then sharpen. If you sharpen a large image and then shrink it, the sharpening is averaged away and you are left with slightly odd edges. Downscaling softens an image, which is exactly why a sharpening pass afterwards is usually worth doing.

Same with cropping and compression: crop, resize, then sharpen, then save. If the image is going to be compressed for the web, sharpen a little less than looks right, because compression exaggerates the artefacts.

What you get back

The output is always a JPEG at high quality, which is the right format for a photograph. A transparent PNG comes back with white where the transparency was, since JPEG cannot store it.

Everything happens on your own device — no upload, no account, no size limit, no watermark.

Related tools

Resize Image is the step to do before this, since downscaling softens an image and creates the need for a sharpening pass. Compress Image comes after — sharpen slightly less if the file is heading for the web. Crop Image reframes first.