How to Compress a Scanned PDF Without Making Text Unreadable
Compressing a digital PDF (such as a document exported directly from Microsoft Word) is easy because the text is stored as clean mathematical vector glyphs. But compressing a scanned PDF is far trickier: the entire page is a single raster photo of paper. If you compress it carelessly, the characters become pixelated, noisy, and unreadable.
The Physics of Scanned Documents
When paper passes through an optical scanner, paper grain, ambient shadows, and ink bleeds are captured alongside the letters. If saved as a standard 300 DPI color JPEG stream, a single contract page can consume 4 MB. When typical cloud compressors attempt to shrink it by 90%, they apply high DCT (discrete cosine transform) quantization, creating "mosquito noise" and artifacts around fine character strokes like the dots on "i"s and thin stems on "t"s.
DPI Sweet-Spot Testing for Scanned Text
What is the optimal resolution for scanned documents? We tested 10-point standard Serif text (Times New Roman) across five DPI levels to measure readability and file weight:
| Resolution (DPI) | Average Page Size | 10pt Text Sharpness | OCR Accuracy | Best Application |
|---|---|---|---|---|
| 300 DPI | 3,200 KB | Crystal Clear | 99.8% | Archival Preservation / Master Copies |
| 150 DPI (Balanced) | 480 KB | Very Sharp | 99.1% | Recommended for 95% of Submissions |
| 110 DPI (High Compression) | 190 KB | Clean & Readable | 97.4% | Government 200 KB Portals |
| 72 DPI (Extreme) | 95 KB | Soft / Minor Artifacts | 91.2% | Strict 100 KB Hard Ceilings |
Three Rules for Clean Scanned Document Compression
- Never downsample below 100 DPI for legal contracts: Signatures, notary seals, and clauses in 8–10 pt fonts need at least 100–110 horizontal and vertical samples per inch to prevent character merging (such as "cl" rendering as "d").
- Convert color scans to grayscale when color conveys no information: Invoices, bank statements, and tax returns rarely need 24-bit RGB. Dropping to 8-bit grayscale strips two-thirds of the color data before compression even begins.
- Use client-side processing for private paperwork: Scanned bank records and signed agreements should never be sent to cloud conversion APIs where employee review or third-party training pipelines could inspect them.
Downsample scanned documents with edge-preserving contrast filters right inside your browser.
About Pushpendra Sharma
AuthorPushpendra is the creator of DocuShrink and a systems engineer specializing in WebAssembly, browser sandboxing, and client-side document processing architectures. He writes extensively on verifiable privacy, byte-stream optimization, and high-performance document tools that run with zero server dependencies.