Extract the images from a document

The real images, not a photograph of the page

When a photo ends up inside a PDF, a Word file or a presentation, the file keeps an image of its own, nearly always larger than you see it on the page: a 4000 by 3000 pixel photo shrunk to five centimetres often stays thousands of pixels wide, because it is the document that draws it small. How many are left depends on the program that saved it, as explained below. This page goes and fetches that object and gives it back to you as it is.

That is the difference from PDF to images, which does another job and does it well: it photographs the whole page, text included, and there the five-centimetre photo becomes the few pixels it takes up in that photograph, recompressed and with the page around it. Here it comes out at the size it has inside the file, on its own and untouched. If you need the page, use that one. If you need the photo, this one.

The real cases are always the same: the supplier's PDF catalogue you need the product photos from, the thesis in Word whose original images you lost, the conference slides with the chart you want to reuse. You can load several files together, even of different kinds, and the images all end up in the same list.

Word, Excel, PowerPoint and LibreOffice: the images come out identical

A .docx, an .xlsx, a .pptx and their LibreOffice cousins (.odt, .ods, .odp) are ZIP archives in disguise: inside there is a folder with the images, saved as files of their own. Here the archive is opened in the browser and the images are taken out byte for byte, without recompressing and without converting them. The archive carries a check code for every file, the CRC: the page recomputes it on every extracted image and, if it does not match, it tells you instead of handing you a damaged file.

What the file does not spell out is the place, and the page rebuilds it by reading the relationships between the parts of the document: the name of each image tells the slide, the sheet or the part where it appears. The slide order is the one of the presentation and not the one of the internal names, which drift apart as soon as someone moves a slide: an image called image7 inside can perfectly well sit on slide 2.

One thing must be said honestly: what comes out is what the program kept. Word and PowerPoint, by default, recompress photos when you insert them and may throw away the cropped parts; if the original photo was 12 megapixels and the document kept 2, here 2 come out. What the file no longer contains cannot be given back. Office vector images (EMF and WMF) come out as they are, but the browser cannot show them as a preview.

The same goes for .epub ebooks, which are ZIP archives too. The thumbnail some programs save inside the file for folder previews is not counted: it is not an image of the document, it is a picture of the first page taken by the program.

In a PDF: JPEGs come out as they are, the others become lossless PNG

Inside a PDF images are objects with a chain of compressions on top. Photographs are nearly always JPEG (in a PDF it is called DCTDecode), and those come out identical to the bytes written in the file: no recompression, no loss, the same fingerprint. If the photo was inserted from a .jpg file, what you download is that file.

The others (charts, screenshots, logos, black-and-white scans) are bare pixels compressed without loss. Here they are decompressed and written back as PNG, which is a lossless format: every pixel has the same value it had in the PDF, including the two colours of one-bit scans, palette images and 16-bit ones. If the image had transparency, which in a PDF sits in a separate image, the page attaches it again as the PNG alpha channel. JPEG 2000 images also come out as they are, as .jp2 or .j2k.

The same image used on thirty pages, like the logo in the header, is a single object in the file and here it comes out only once, with a note of the pages it appears on. It works across different files too: if you load five quotes with the same logo, you find it once, with the list of files.

What I cannot extract I say, I do not make it vanish

Some PDF images are in formats the browser cannot decode. Black-and-white scans made by photocopiers and office scanners often use JBIG2 or fax compression; images meant for print can be in CMYK without JPEG or in special colours. Pulling them out would mean inventing the colours, and a wrong image presented as right is worse than a no: so they appear in a separate list, with the reason and the page.

Then there are the images placed inline, small pictures written straight into the page content instead of as an object of their own: they are usually icons and bullets. They do not come out here, but the page counts which pages have them and says so.

If a PDF is password protected, even only with printing or copying restrictions, the images inside are encrypted: the page recognises it and does not hand over meaningless bytes. The protection has to be removed first with Remove a PDF password, and for a protected Word or Excel file with Remove the password from an Office file. The old Office 97-2003 .doc or .ppt format is a different thing and does not open here: just save it as .docx or .pptx.

A case that surprises people: a PDF edited several times can contain images that appear on no page, leftovers of an earlier version that the program did not clean up. Here those come out too, marked as such: it may be exactly the photo somebody thought they had removed.

Downloading them, and what they are called

Every image has its own card with the preview, the size in pixels, the weight and the format, and it downloads on its own with one tap. With the button at the top you take them all together in a ZIP file, which does not recompress them: they stay identical in there too.

The names say where they come from: the document name, then the page or the part, then the number. For example catalogue-page07-1.jpg is the first image found on page 7 of the catalogue, and report-header-image1.png is the logo in the header of the Word file. So, even after putting them all in one folder, you know where each one came from.

What it does not do

It does not turn vector drawings into images. A chart made in Excel, a logo drawn with paths or a table are not images inside the file: they are drawing instructions, and there is nothing to extract here. If you need the page as you see it, with everything on it, use PDF to images.

It does not recompress and does not convert. If an extracted image is too heavy, Compress image shrinks it; if the problem is the weight of the whole document, Compress an Office file recompresses the photos leaving the text as it is.

It does not read the text in the images. To copy the words from a scan or a screenshot there is Image to text (OCR), and to see the date, camera and location saved inside a photo there is Photo EXIF viewer: JPEGs come out identical, so that data, if it was there, is still there.