Extract Images from PDF

Every image sitting inside a PDF is stored as its own separate object, referenced from whichever page displays it — a scanned photo, a product picture, a logo in a header, a diagram pasted in from somewhere else. Unlike converting a PDF page into a picture, which flattens everything on that page into one new image, extraction reaches into the file's actual structure and pulls out exactly those original embedded objects, at exactly the resolution they were embedded with.

This guide covers how that extraction actually happens, the real difference between a raster image and the vector graphics a PDF can also contain, what the filtering options are doing behind the scenes, and where pulling images back out of a PDF is genuinely the fastest way to get a job done. None of it requires knowing anything about how PDFs are built internally — just enough to know what to expect from the results.

How the Extraction Actually Works

Every PDF page is described by a list of drawing instructions — draw this line, fill this shape, place this image at these coordinates. This tool reads that instruction list for every page and watches specifically for the commands that place an embedded image, rather than the commands that draw text or vector shapes.

When one of those image-placing instructions is found, it points to the actual image object stored inside the file — its raw pixel data, at its original width and height. That data is decoded and painted onto a blank canvas at those exact native dimensions, then saved out as a JPG. Nothing about the page layout, the surrounding text, or the rest of the document is involved at any point; only the image data itself is touched, page by page, until every page has been checked.

Raster Images vs. Vector Graphics

Not everything that looks like a picture in a PDF is stored as one. A photo, a scan, or a picture pasted in from another program is almost always a raster image — a grid of individual pixel values, exactly the kind of object this tool extracts. A logo, chart, or diagram, on the other hand, is very often drawn using vector instructions instead: a set of lines, curves, and fill commands that get rendered fresh every time the page is displayed, with no pixel grid stored anywhere in the file.

Vector graphics genuinely aren't images in the file-format sense, which is why a diagram or chart sometimes doesn't show up in extracted results even though it's clearly visible on the page. There's no raster data sitting there to pull out — the shape only exists as a set of drawing instructions, the same category as the text on the page rather than the photo beside it.

What the Filtering Options Actually Do

Two settings sit above the extract button, both turned on by default, both aimed at the same goal: getting back the photos someone actually wants, not a folder cluttered with page furniture.

Skip tiny icons and decorations filters out anything below a small pixel-dimension threshold, since these are almost always bullet graphics, small logos, or decorative dividers rather than actual photos — the kind of thing that technically counts as an embedded raster image but was never what anyone meant by "extract the images."

Skip repeated images compares each extracted image's exact data against everything already found, and silently drops anything that's a byte-for-byte match — the same company letterhead appearing on all thirty pages of a report becomes one saved file instead of thirty identical copies. Both filters can be turned off individually if a specific job genuinely needs every last embedded object, tiny or repeated.

Downloading One Image vs. All of Them

For a PDF with just one or two photos worth keeping, downloading them individually from the gallery is the quicker path. For a document with a dozen or a hundred embedded images — a scanned photo album turned into a PDF, or a product catalogue — the "Download All as ZIP" option bundles every extracted image into a single archive in one click, avoiding a hundred separate browser downloads landing in the same folder.

The ZIP is assembled the same way the images were extracted: entirely in your browser, with no intermediate upload to a server to build the archive.

Privacy for Documents Full of Photos

PDFs packed with embedded images are frequently the exact kind of document worth being careful with — scanned ID booklets, medical imaging reports, property listings with interior photos, an insurance claim's photo evidence. Reading the file, decoding its images, and building the ZIP archive all happen inside your browser using JavaScript, so none of that content needs to reach a server to come back out as individual files.

You can confirm this directly rather than taking it on trust: open your browser's developer tools, watch the Network tab while extracting, and see that no request carrying your file or its images goes out at any point.

Common Reasons to Extract

• Real estate listings. Pulling interior and exterior photos back out of a PDF brochure to reuse on a listing site or in a new flyer.
• Recovering photos from an old scanned album. A PDF someone made years ago from a stack of printed photos, with the original loose images long since lost.
• Product catalogues. Getting individual product photos out of a supplier's PDF catalogue for use on a storefront.
• Academic and research use. Pulling figures, photographs, or diagrams out of a paper for citation or further analysis, where the original image files were never separately provided.
• Insurance and legal documentation. Extracting photo evidence attached to a claim or case file into individual images for a separate review process.
• Design and marketing handoffs. Recovering source images from a client-approved PDF proof when the original design files aren't available.
• Genealogy and family archives. Pulling individual photographs back out of a PDF someone assembled from a scanned family album years ago.
• Journalism and publishing. Recovering original photographs from a press kit or archived PDF layout for reuse in a new article.

Troubleshooting

No images were found at all. The PDF's pictures may be vector graphics rather than raster images, or the file itself may be a single scanned page where the "image" is actually the entire page background — try extracting with the size filter turned off to confirm nothing was missed.

An image extracted looks stretched or distorted. This is rare and usually means the original PDF itself applied a non-uniform scale to that image when placing it on the page; the extracted file reflects the image's native pixel data, independent of how it happened to be stretched or compressed visually on the page.

Fewer images came out than expected. Check whether deduplication removed repeated copies of the same picture, or whether the size filter is skipping something that's technically small but still wanted — both are adjustable before running the extraction again.

One specific image failed to extract while the rest worked fine. A small number of unusually encoded images can occasionally fail to decode; the tool skips that one image and continues with the rest rather than stopping the whole extraction.

Extraction seems slow on a large PDF. A document with many high-resolution images takes proportionally longer to scan and decode, since every page's instruction list has to be read individually — this scales with your device's own processing speed rather than a queue or upload time.

Frequently Asked Questions

How Duplicate Detection Actually Compares Images

Checking whether two images are "the same" by comparing every pixel would be slow across a large document, so this tool instead computes a short cryptographic fingerprint of each extracted image's data — the same category of technique used to verify downloaded files haven't been corrupted or tampered with.

Two images produce the identical fingerprint only if their underlying data matches exactly, byte for byte. This means a logo that's been recompressed, resized, or even saved at a slightly different quality between two placements in the same document won't be caught as a duplicate — only genuinely identical embedded copies are. For a letterhead or watermark placed the same way on every page, which is the common real-world case, this catches every repeat cleanly.