Sharpen Image

Sharpening exaggerates differences between neighbouring pixels. This tool uses a 3 x 3 convolution: for each pixel, it takes the pixel's own value multiplied by 1 + 4k and subtracts k times each of the four pixels directly above, below, left and right, where k is the intensity divided by 50. On flat areas the neighbours cancel out and nothing changes; along an edge, the light side gets lighter and the dark side darker, so the edge looks crisper.

At the default intensity of 50, k is 1 and the filter is the standard textbook sharpening kernel. At 100, k is 2 and the effect is twice as strong. Despite the address of this page, no AI is involved; it is a fixed mathematical filter applied to every pixel.

• Images that will be shown smaller than their pixel size often need less sharpening than you think, because downscaling hides softness.
• Photos for the web usually look best with low to moderate values. Zoom to 100% to judge, not the scaled preview.
• Graphics and text can take more, but anti-aliased edges may start to look jagged.

Watch for halos (thin light or dark outlines along edges) and for noise turning into visible grain. Both mean the setting is too high. If the photo is noisy to begin with, consider the Denoiser first and a light sharpen afterwards.

Sharpen last, after resizing and cropping. Resizing changes the scale of edges, so sharpening done before a resize is partly undone or exaggerated by it. A practical sequence is: crop, resize to the final size with the Image Resizer, sharpen here, and finally compress if needed.

The filter runs in JavaScript over every pixel, so large photos can take several seconds. The result is saved as WebP at quality 0.9 at the original dimensions, without EXIF data. For blurry rather than slightly soft photos, read the Unblur Image page, which explains what sharpening can and cannot fix.

All processing is done in your browser. The image is not uploaded.