Photo editing guides

Clean a speckled scan while keeping small lettering

·

A bright speckle in flat paper and faint grain across a scan call for different tests. Use the built-in example to compare both filters while keeping the source and crop fixed.

Image Noise Reducer

Remove an isolated paper speckle

Load the example and select Median at radius 1. Reduce noise. At the example pixel (20, 20), using zero-based coordinates, one bright gray value of 248 is surrounded by eight values of 144. The middle value of those nine is 144, so the spot matches the nearby paper.

SciPy median filter reference

Compare Gaussian on the same paper

Switch to Gaussian at radius 2 and run the filter again. The same spot gets dimmer, and the near neighbors receive part of its brightness. Examine a dark letter in both previews to see how this same averaging softens a hard edge.

SciPy Gaussian filter reference

Check fine marks in the saved file

Choose the smallest radius that makes the unwanted marks less visible. Compare a thin letter, a flat patch and any transparent edges. Save PNG if you want to inspect the filtered values without lossy compression.

W3C Portable Network Graphics specification

Common questions

Why did Gaussian leave a pale halo around a speckle?

Gaussian averaging spreads part of a bright pixel's contribution across nearby pixels. Try Median at radius 1 for isolated extremes in a flat patch.

Will a large Median radius remove fine marks?

It can remove features that cover less than half of a neighborhood. Compare small punctuation and hairline strokes before using a larger radius.

Sources

Put it into practice.

Related guides