Why Some Scanned PDFs Will Not Compress

Why Some Scanned PDFs Will Not Compress

The October 2026 audit of FlashKit Tools found a 6-page, 19 MB “scan” that saved 0% at every quality. The JPEGs were real. They sat inside Form XObjects with a filter array [ASCII85Decode, DCTDecode]. The old compressor looked only for a single /DCTDecode name and then tried to feed ASCII85 text to a JPEG decoder. Nothing matched. That path is fixed. This page is the honest list of what still will not move.

Run the file on PDF compressor and read the notes. Related: 200 KB target and 150/200 DPI.

Recompress a stubborn scan

Forms and ASCII85+DCT are included now. JBIG2/CCITT still are not. The notes say which case you hit.

Open PDF Compressor

What we rewrite now

  • JPEG (DCTDecode) — recompressed and optionally downsampled from placed size.
  • Filter arrays — ASCII85 or ASCIIHex wrappers are decoded first; the JPEG is written back as DCT only.
  • Form XObjects — nested images are found by walking page content and form resources.
  • Flate RGB / DeviceGray — the old “PNG untouched” case. Inflated, predictor undone when present, then JPEG if that is smaller.

What still saves 0%

  • JBIG2 and CCITT (G4) — bitonal fax encodings. Already tiny per page relative to a colour JPEG, and not something a canvas JPEG encoder should pretend to improve.
  • Vectors, text, and fonts — there is no photo to downsample. Object-stream rewrite may shave a few KB or add a few.
  • Already-optimized JPEGs — if 40% quality is still larger than the original stream, the original is kept. Shrinking then requires DPI, not another quality pass.
  • JPEG2000 (JPX), 1-bit ImageMasks, soft masks — skipped so a transparency mask does not get out of sync with a resized base image.
  • Encrypted PDFs — rejected. Unlock a copy in a desktop reader; FlashKit Tools will not ask for a password.

How to read the notes

“Recompressed N JPEG images” means the DCT path ran. “Converted N PNG/Flate images to JPEG” is the old PNG hole. “Unwrapped ASCII85” is the filter-array fix. “Walked Form XObjects” means nested forms were not skipped. “No embedded images were smaller” is the honest 0% case — try Split PDF if extra pages are optional, or rebuild from photos with JPG to PDF.

If the tool did recompress and you still cannot read the page, you went too far on DPI. That is a different failure than 0%. Raise the target or pick 150/200 DPI in Quality mode.

Related Tools