Remove tiny colourful fibers (not dust) from scanned image

In GIMP (or equivalent), you could:

1. Convert image to 1-bit black & white (Image > Mode > Indexed...)
2. Convert image back to RGB (Image > Mode > RGB)
3. Apply some low-pass filter to smooth out the edges (Filters > Blur > ...)

Here's the result with 1 iteration of the Mean Curvature Blur filter in step 3, applied directly to the raw scan:

It's not perfect, but perhaps useful as a first approximation.

Any filter that is able to remove dust and fibers will risk affecting fine details you wish to preserve. Some options to limit unwanted effects:

• Manual masking. Select each unwanted speck manually. Time consuming.
• Edge detection. Fine details are still at risk, but the mask can be manually refined.
• Wavelet decompose. Details of different sizes are confined to their own layer. Changes to one layer are prevented from affecting others.

Rather than attempt to remove all specks at once, you can apply multiple methods at reduced strength so that desired details are preserved. Different methods will remove different specks, and any that remain can be addressed manually.

Rather than attempt to blur the specks away, I would prefer methods that create masks that can be filled by in-painting algorithms. (Similar to using a healing brush.) The results would be "cleaner". There would be no left over gray haze from blurring.