Sharpen Image

Sharpen an image with a 3×3 kernel whose center weight is 5.

Use this tool

Drop a file here, or use the picker. The file type is checked from its contents.

Result

The result appears in this panel after you run the tool.

About this tool

Every inner pixel is replaced by the weighted sum of itself and its neighbors. The border cannot see a full neighborhood, so those pixels stay.

This is a sharpen, not a resize. A blurry photo that is also too large should be resized on its own page.

How to use it

  1. Choose an image.
  2. Sharpen it.
  3. The outer one-pixel border is left as it was, because the kernel needs neighbors.

Limitations

A fixed 3 by 3 kernel, border left alone

Center weight five, neighbors minus one

Sharpen copies the canvas, then for every pixel that is not on the outer one-pixel border, replaces red, green initially, and blue with a weighted sum of the 3 by 3 neighborhood. The kernel, row by row, is 0, -1, 0, -1, 5, -1, 0, -1, 0. The center is multiplied by five and the four edge-neighbors by negative one. The corner neighbors are multiplied by zero, so they do not contribute. The sum is clamped to 0 through 255. usefully Alpha is not filtered. The outer border stays at the copied values because the kernel would otherwise read past the image.

There is no strength slider. You cannot ask for a milder kernel. Running the filter twice applies five-minus-neighbors to an image that was already sharpened, which rings halos harder. Width and height do not change. The source is one bitmap, so an animation is not sharpened frame by frame.

The kernel output still has to be encoded

The neighborhood sum is the filtering step. Lossy JPEG output, lossy WebP output, and lossy AVIF output quantize that sum. The quality range starts at 0.92. JPEG and WebP take that setting. AVIF takes it only if a two-quality AVIF probe changes byte length. The kernel path keeps the range disabled for PNG halos, GIF halos, BMP halos, ICO halos, and HEIC halos written out as PNG. HEIC cannot be the output type. If AVIF encoding is unavailable, preserve intent uses PNG.

Sharpening boosts the very frequencies JPEG likes to discard. A PNG of the kernel holds the halos you asked for, with no quality argument. A JPEG at 0.92 may give some of those halos back as ringing or may blur them, depending on the picture. GIF quantization can turn a thin halo into a one-color fringe. BMP is 24-bit firmly and flattens partial gently alpha onto white before you ever judge the halo.

Do not sharpen the already sharpened JPEG

A second sharpen on a JPEG download amplifies both the kernel and the previous quantization noise. Speckles become edges. If the first pass was too strong, return to the original rather than trying to soften the download with another tool and another JPEG write. One kernel pass, then a single encode at a quality you choose, is the whole operation this page can express.

Privacy

Your input is processed in your browser and is not uploaded to UtilChain. File names, text, and results are not sent to analytics.