How to Redact a Photo or Image Embedded Inside a PDF
Photos and other images inside a PDF are not text runs. You redact them by destroying or rewriting the pixels in that region — not by selecting letters with a text tool.
Many PDFs mix selectable text with embedded images: a passport photo on an application, a screenshot pasted into a brief, a logo that includes a face, a phone camera picture of a whiteboard dropped onto page three. Text-selection redaction is the wrong mental model for those regions. There is no character string for “find face” to match. The sensitive material lives in image samples — pixels — referenced by the page. Redacting them means covering that area and exporting in a way that the original pixel data in the covered region is gone, not merely hidden under a rectangle someone can peel off or an image object someone can extract whole.
This is why image redaction feels different from highlighting a Social Security number in a paragraph. For text, tools can target operators and fonts. For photos, the durable fix is geometric: mark the rectangle (or rectangles) that contain the sensitive imagery and burn the page so those pixels are overwritten or the image is replaced with a version that no longer contains them. Visual-only sticky notes and comment shapes that sit above the image without rewriting it are the classic failure mode.
What to redact (pixel-level vs text-layer)
Identify whether the thing you need to hide is truly an image object or text that only looks photographic. A scanned full-page PDF is often one large image per page (sometimes with a hidden OCR text layer on top). A born-digital form may be mostly text with a small JPEG of an ID photo in the corner. Screenshots may be images that contain readable UI text as pixels only. Each case still uses box-based coverage for the image portion, but you should also check for parallel text: labels beside the photo, OCR under a scan, or captions that repeat what the image shows.
Permanent redaction of an embedded photo aims to destroy the covered samples in the shared file. That may rewrite a subset of a larger image, replace the image resource, or flatten the page into a new image with black fill where you marked. Visual redaction that only draws an annotation leaves the XObject image intact; a determined recipient can open the PDF in an editor and export the original photo. The distinction between permanent and visual redaction matters even more for images than for text because people assume “I can’t select it, so it’s safe,” forgetting that images are not selected as text in the first place.
Scanned-versus-native workflows change the packing of images but not the rule: if sensitive content is pictorial, mark the pixels. Native PDFs may embed many small images; scanned PDFs may embed few huge ones. Guides that compare those formats help you predict file-size swings after burn-in, but the marking gesture is the same — draw over what must disappear.
How to do it
Open the PDF in a local redaction tool. With HidePDF, drop the file into the browser tool — your file never leaves your device — navigate to the page with the photo, and draw a black box that fully covers the sensitive imagery. Include edges, reflections in badge plastic, and any second inset thumbnail of the same photo. If a document number is printed on the image itself, cover that region too; it will not appear in text search.
Download the burned-in export. Re-open it and zoom into the former photo area. You should see opaque coverage with no recoverable face or ID details in that rectangle. Try tools you would normally use to extract images from PDFs on a spare copy of the export; the covered region should not yield the original portrait. If you only added a comment square and the portrait still extracts, redo the pass with a permanent burn-in. Also search the file for any typed duplicates of data visible in the photo (names, ID numbers) and redact those text runs separately.
When the whole page is a scan, treat the photo region like any other area on that scan: box it, burn it, then consider whether OCR text still names what you covered. When multiple pages reuse the same image resource, confirm each placement is covered or that the shared resource was actually rewritten — depending on your tool’s behavior, one mark may not automatically fix every reuse. Spot-check every page that shows the image.
Realistic scenarios
HR redacts employee badge photos from onboarding PDFs before sharing with a vendor. The names in the text layer were already handled; the portraits were not, and a visual black square left extractable JPEGs. A school releases discipline records and covers a student’s face in a hallway camera still pasted into the packet; the still is an embedded PNG, so staff draw boxes over the face and burn in rather than trying to “select” the face. A lawyer receives a PDF with a photographed whiteboard containing a phone number; the number exists only as pixels in the photo, so text find fails and manual image coverage is required.
A clinic shares a referral PDF that includes an insurance card photo. Cropping the page in a word processor and reprinting to PDF may still leave fragments or alternate image objects. Box redaction on the PDF itself is clearer. An investigator redacts a map screenshot; street labels inside the image are pixels, while the figure caption below is text — both need attention.
Someone exports the embedded photo, paints over it in an image editor, and reinserts the wrong resolution or forgets a second copy on page five. In-place PDF redaction with verification across pages avoids that shuffle when a capable burn-in tool is available.
Common mistakes
Relying on text search to find faces or ID cards. Using annotation covers that leave the image XObject intact. Covering only the eyes while leaving the rest of a unique badge photo identifiable. Forgetting captions, EXIF-like metadata in some workflows, or duplicate thumbnails. Assuming a scanned page has no text layer after you black out part of the image. Confusing “I flattened the PDF” with “I destroyed the photo samples.”
Another mistake is redacting a low-resolution preview on screen with a tiny box that misses hair, ears, or card edges at print resolution. Zoom in before you finalize marks. Finally, do not treat passwording the PDF as image redaction; the photo is still inside for anyone who can open the file.
Related guides
Explore more ways to redact PDFs privately, or use the redaction tool above:
- Redact a scanned PDF vs a native PDF
- Permanent PDF redaction vs visual redaction
- Why a QR code or barcode can undo your PDF redaction
Frequently asked questions
Can I redact an embedded photo the same way I redact text?
You still draw a region and burn it in, but the mechanism targets image pixels (or the image object), not a text run. Text-based find tools will not locate faces or ID numbers that exist only inside a photo.
Is placing a black box on top of a photo enough?
Not if the original image object remains underneath and can be extracted. Use a permanent redaction export that removes or rewrites the covered image data, then verify the photo region cannot be recovered.
What if the PDF page is entirely one big scan?
Treat the sensitive area as part of that page image: draw boxes over the pixels that must go and burn in so those pixels are destroyed in the export, remembering there may also be a separate hidden text layer from OCR.
Do I need a separate image editor to redact a photo inside a PDF?
Usually no. A PDF redaction tool that burns boxes into the page can cover embedded photos in place. Exporting the image, editing it elsewhere, and reinserting it is slower and easy to get wrong.