PDF Compression Techniques Explained

1. Object Cleanup

What it does: Removes objects that no page actually references - old thumbnails, leftover edit data, duplicate fonts, broken bookmarks.

Best for: Any PDF that has been edited or re-saved multiple times in Word, Acrobat, or InDesign.

Typical savings: 5-20%. Sometimes much more on heavily edited documents.

Quality cost: None. This is purely garbage collection.

2. Stream Re-Filtering (Flate / Deflate)

What it does: Wraps every text and data stream in the Flate filter (the same algorithm used by zip and PNG). Many older PDFs leave streams uncompressed entirely.

Best for: Older PDFs, PDFs generated by simple printer drivers, and anything created before 2010.

Typical savings: 10-40% on uncompressed input.

Quality cost: None. Flate is lossless.

3. Image Downsampling

What it does: Reduces image resolution to match how the image is actually used on the page. A 4000-pixel-wide photo placed in a 600-pixel-wide column gets resampled down to ~600 pixels.

Best for: PDFs exported from cameras, phones, or design tools where every photo is at full resolution.

Typical savings: 50-90% per image.

Quality cost: Invisible at normal screen zoom. Visible only when zooming far past 100%.

4. JPEG (DCT) Re-Encoding for Photos

What it does: Photos and continuous-tone images are re-encoded as JPEG with an adjustable quality setting (typically 60-85 for general use).

Best for: Brochures, portfolios, scanned color documents.

Typical savings: 30-70% per photo.

Quality cost: Slight, mostly in subtle gradients. Most viewers will not notice.

5. JPEG2000 for High-Quality Images

What it does: A more advanced image codec that produces smaller files than JPEG at the same visual quality.

Best for: High-end print documents, art portfolios where quality matters.

Typical savings: 20-30% better than JPEG at the same quality target.

Quality cost: Lower than JPEG. Downsides are slower decode and weaker viewer support on very old PDF readers.

6. CCITT Group 4 for Black-and-White Scans

What it does: A lossless encoding designed specifically for 1-bit black-and-white images. Originally invented for fax machines.

Best for: Pure B&W scanned forms, contracts, line drawings.

Typical savings: Massive - often 95%+ vs storing the same image as raw bitmap.

Quality cost: None. Lossless on 1-bit images.

7. JBIG2 for B&W Document Scans

What it does: The most aggressive method for black-and-white scans. JBIG2 detects repeating shapes (like the letter "e") and stores each one only once.

Best for: Long scanned books, archives, government records.

Typical savings: 2-5x smaller than CCITT.

Quality cost: If used in lossy mode, similar-looking characters can be confused. Use lossless JBIG2 for legal documents.

8. Font Subsetting

What it does: Embeds only the glyphs your document actually uses. A font that ships with 5,000 glyphs typically gets pruned to 50-200 in real documents.

Best for: Documents with many embedded fonts - design exports, multilingual reports.

Typical savings: 20-80% per font.

Quality cost: None for typical use. Editing the PDF later may show "missing glyph" boxes if you add characters not in the subset.

9. Object Stream Compression

What it does: Bundles many small objects into a single compressed stream rather than storing each separately. Available in PDF 1.5+.

Best for: Documents with thousands of small objects - tagged PDFs, accessibility-rich documents, complex forms.

Typical savings: 5-15%.

Quality cost: None.

10. Cross-Reference Stream Compression

What it does: Replaces the verbose ASCII cross-reference table at the end of older PDFs with a compact compressed stream.

Best for: Any PDF still using the legacy table format.

Typical savings: 1-5%. Small but free.

Quality cost: None.

Quick Reference: Which Technique for Which PDF