Can I really compress a photo PDF without ANY visible quality loss?

Yes, but only to a point. "Lossless" compression on already-compressed photos (which is what most embedded PDF images are) saves at most 5–15% in file size. That said, "no visible quality loss" is a different bar from "byte-identical lossless" — JPEG quality 90 is visually indistinguishable from quality 100 for the human eye, but produces files 40–60% smaller. The right framing is "preserve perceptual quality" rather than "preserve every pixel"; for photos, the perceptual definition is what matters.

When should I use JPEG vs PNG for embedded PDF images?

JPEG for photographs and continuous-tone images (faces, landscapes, product photos) — JPEG's discrete cosine transform is tuned for the statistical properties of natural images. PNG for line art, screenshots, diagrams, logos, anything with flat colour regions or sharp edges (text, vector graphics rasterised to pixels). The reason: JPEG produces ringing artefacts around sharp edges (visible halos around text); PNG's lossless compression keeps those edges sharp. For a PDF that contains both content types, the right answer is per-image format selection rather than one-size-fits-all.

What about JPEG 2000?

JPEG 2000 (wavelet-based, JP2) compresses about 20–30% better than JPEG at equivalent visual quality and supports a lossless mode that beats PNG on photographic content. Native PDF support has existed since PDF 1.5 (2003). The catch: some older readers and most printer drivers handle JPEG 2000 poorly, with rendering glitches or outright failures. Use for archival PDFs where modern readers are guaranteed; avoid for documents that will be printed or distributed to mixed environments.

My PDF is mostly text but file size is huge — why?

Almost always: a few high-resolution embedded images you did not realise were in the document. A 1-page report with a logo at 300 dpi is 1.5 MB; the same page with the logo at 96 dpi (still crisp for screen) is 150 KB. Other common culprits: embedded full-character fonts (instead of subsetted), an embedded copy of the source document, or hidden full-page background images. The compress tool reports per-page byte breakdown so you can spot the outliers.

Will my photos go through a generation-loss cycle if I re-compress?

Yes if you re-encode JPEGs at lower quality each time. JPEG is lossy; every encode-decode-encode cycle accumulates artefacts. The first re-encode (from JPEG 100 to JPEG 90) is invisible; the second (to 80) starts to show subtly; by the fifth (to 60) artefacts are obvious. Practical guidance: do compression at the SOURCE (once, with the highest-quality original) and stop re-encoding the same image. If you must compress an already-compressed PDF, pick a relatively gentle target (quality 80+) to slow the generation-loss accumulation.

Is my PDF uploaded to your servers during compression?

No. The compression runs entirely in your browser using pdf-lib for PDF surgery and the browser's canvas API for image re-encoding. Your file is loaded into a sandboxed memory buffer, each embedded image is decoded, re-encoded at the chosen settings, and written back. Verify in DevTools Network — zero outbound requests. Important for photo-heavy PDFs that contain sensitive content (medical images, financial documentation, personal documents).

How can I tell if compression hurt quality without doing the full A/B comparison?

Use the "Visual diff" mode after compression. The tool overlays the before / after at 100% zoom and highlights pixel differences. For text-heavy regions, look for blocking artefacts (8×8 pixel squares) around character edges; for photos, look for ringing (halos around sharp edges) and smooth-region banding. If the diff highlight shows mostly noise across photographic regions and no structured artefacts around text, the compression is safe. If you see structured artefacts (squares, halos, bands), back off the quality setting and retry.

Compress PDF Photos Without Quality Loss

6 min read

Compress PDF photos without quality loss — JPEG vs PNG tradeoffs

By ScoutMyTool Editorial Team · Last updated: 2026-05-20

After working with hundreds of users on photo-heavy PDF compression cases, the recurring confusion is that "PDF compression" lumps together two operations that work very differently: shrinking the structural overhead of the PDF (fonts, cross-reference, object streams) and re-encoding the embedded raster images. The first saves a few percent; the second can save 50–80% on a photo-heavy file, but only if you pick the right encoder for each kind of image content. Below is the format decision tree, the quality settings that actually matter, and the workflow that holds quality while cutting file size.

Format choice by image content

Content type	Best format	Quality	Notes
Photographs (faces, landscapes, products)	JPEG	Q85–Q90	Best size/quality tradeoff for natural images.
Screenshots of text or UIs	PNG	Lossless	Sharp edges; small file size on flat regions.
Line art, diagrams, logos	PNG	Lossless	No JPEG ringing on sharp edges.
Charts with photo backgrounds	JPEG (Q90)	Q90	JPEG handles mixed content acceptably at Q90.
Archival photographs	JPEG 2000 lossless OR PNG	Lossless	JPEG 2000 saves ~20% over PNG on photos.
Scanned text-heavy pages	JBIG2 (bitonal) OR PNG	Lossless	JBIG2 is the standard for scanned documents.
Mixed text + photo pages	Hybrid (text PNG, photo JPEG)	Per-region	Best compression; tool decides per region.

Step-by-step: compress a photo-heavy PDF without visible quality loss

The ScoutMyTool tool lives at scoutmytool.com/pdf/compress-pdf. Runs client-side — no upload, no signup, no quota.

Drop your photo-heavy PDF. Loads into a sandboxed memory buffer; nothing is uploaded.
Review the per-page byte breakdown.The pre-compress report shows which pages and which images are dominating file size. Almost always a small number of high-resolution photos account for the bulk.
Pick "Smart compression". This mode applies per-image format selection: photographs go to JPEG at Q85; line art and screenshots go to PNG lossless; bitonal scans go to JBIG2. Right default for mixed-content PDFs.
Or pick a fixed-format mode. "All JPEG at Q90" for pure photo PDFs (max compression). "All PNG lossless" for documents that must not lose any pixel (legal exhibits, archival photos).
Pick downsampling target. Default 150 dpi (federal accessibility minimum, sharp on screen). Drop to 96 dpi for "online-only" documents where file size dominates. Keep at original resolution for print-bound files.
Click Compress. Each embedded image is decoded, re-encoded at the chosen format and quality, and written back into the PDF. Progress shown live.
Use Visual Diff to verify quality.Toggle "Show visual diff" — the tool overlays before/after with pixel differences highlighted. Look for structured artefacts (blocks around text, halos at edges). Random pixel-level noise on photographs is fine; structured artefacts are not.
Download. The compressed PDF is delivered as a download. Compare file sizes before/after; for a photo-heavy file at smart compression, expect 50–75% reduction.

The format theory in one short read

Three families of image compression matter for PDF:

JPEG (ISO/IEC 10918) — lossy, DCT-based, tuned for continuous-tone images¹. Excellent for photographs at quality 75–95. Produces visible blocking and ringing at quality <60. No alpha channel.
PNG (W3C / ISO/IEC 15948) — lossless, DEFLATE-based, with alpha channel². Compresses flat regions and sharp edges well; poor on photographs (where JPEG's lossy approach is the right tradeoff).
JPEG 2000 (ISO/IEC 15444) — wavelet- based, supports both lossy and lossless modes, compresses 20–30% better than JPEG at equivalent visual quality. Less universally supported by readers and printers.

The PDF specification (ISO 32000-1) defines image filters for each of these (DCTDecode for JPEG, FlateDecode + bit-depth handling for PNG, JPXDecode for JPEG 2000) along with JBIG2Decode for bitonal scans and RunLengthDecode for very simple images³.

Frequently asked questions

Can I really compress a photo PDF without ANY visible quality loss?: Yes, but only to a point. "Lossless" compression on already-compressed photos (which is what most embedded PDF images are) saves at most 5–15% in file size. That said, "no visible quality loss" is a different bar from "byte-identical lossless" — JPEG quality 90 is visually indistinguishable from quality 100 for the human eye, but produces files 40–60% smaller. The right framing is "preserve perceptual quality" rather than "preserve every pixel"; for photos, the perceptual definition is what matters.
When should I use JPEG vs PNG for embedded PDF images?: JPEG for photographs and continuous-tone images (faces, landscapes, product photos) — JPEG's discrete cosine transform is tuned for the statistical properties of natural images. PNG for line art, screenshots, diagrams, logos, anything with flat colour regions or sharp edges (text, vector graphics rasterised to pixels). The reason: JPEG produces ringing artefacts around sharp edges (visible halos around text); PNG's lossless compression keeps those edges sharp. For a PDF that contains both content types, the right answer is per-image format selection rather than one-size-fits-all.
What about JPEG 2000?: JPEG 2000 (wavelet-based, JP2) compresses about 20–30% better than JPEG at equivalent visual quality and supports a lossless mode that beats PNG on photographic content. Native PDF support has existed since PDF 1.5 (2003). The catch: some older readers and most printer drivers handle JPEG 2000 poorly, with rendering glitches or outright failures. Use for archival PDFs where modern readers are guaranteed; avoid for documents that will be printed or distributed to mixed environments.
My PDF is mostly text but file size is huge — why?: Almost always: a few high-resolution embedded images you did not realise were in the document. A 1-page report with a logo at 300 dpi is 1.5 MB; the same page with the logo at 96 dpi (still crisp for screen) is 150 KB. Other common culprits: embedded full-character fonts (instead of subsetted), an embedded copy of the source document, or hidden full-page background images. The compress tool reports per-page byte breakdown so you can spot the outliers.
Will my photos go through a generation-loss cycle if I re-compress?: Yes if you re-encode JPEGs at lower quality each time. JPEG is lossy; every encode-decode-encode cycle accumulates artefacts. The first re-encode (from JPEG 100 to JPEG 90) is invisible; the second (to 80) starts to show subtly; by the fifth (to 60) artefacts are obvious. Practical guidance: do compression at the SOURCE (once, with the highest-quality original) and stop re-encoding the same image. If you must compress an already-compressed PDF, pick a relatively gentle target (quality 80+) to slow the generation-loss accumulation.
Is my PDF uploaded to your servers during compression?: No. The compression runs entirely in your browser using pdf-lib for PDF surgery and the browser's canvas API for image re-encoding. Your file is loaded into a sandboxed memory buffer, each embedded image is decoded, re-encoded at the chosen settings, and written back. Verify in DevTools Network — zero outbound requests. Important for photo-heavy PDFs that contain sensitive content (medical images, financial documentation, personal documents).
How can I tell if compression hurt quality without doing the full A/B comparison?: Use the "Visual diff" mode after compression. The tool overlays the before / after at 100% zoom and highlights pixel differences. For text-heavy regions, look for blocking artefacts (8×8 pixel squares) around character edges; for photos, look for ringing (halos around sharp edges) and smooth-region banding. If the diff highlight shows mostly noise across photographic regions and no structured artefacts around text, the compression is safe. If you see structured artefacts (squares, halos, bands), back off the quality setting and retry.

Compress your photo-heavy PDF now — free, no signup

Smart per-image format selection, downsampling control, visual-diff verification. Runs entirely in your browser.

Open the Compress PDF tool at scoutmytool.com/pdf/compress-pdf →