The same request, "save the pictures from this PDF", can mean two different jobs. One is to pull out the photos that were placed in the document. The other is to turn each page into a picture. They give different files, and mixing them up is the main reason people end up with blurry, tiny or oddly cut images.
Extracting is not the same as converting
Extracting means finding the image objects stored inside the PDF and saving them. Each one has its own width and height in pixels, set when the PDF was made. A 4000 pixel wide photo placed in a brochure at 8 cm wide is still 4000 pixels wide inside the file. Extraction can return all of them.
Converting means rendering a page, as a viewer would show it, onto a canvas of a chosen resolution and saving that canvas. It includes text, vector drawings and every picture on the page, flattened into one image. The pixel size depends on the dpi you choose, not on the original photo. The same 8 cm photo at 300 dpi is about 945 pixels wide (8 cm is 3.15 in, times 300). That is a fraction of the 4000 pixels in the file.
- Need the original photo, logo or chart, as sharp as the PDF holds it: extract.
- Need a whole page as a picture, with the text and layout, to post or paste: convert.
- Need a picture of a vector drawing or a chart drawn with lines: convert. Drawn graphics are not stored as images, so there is nothing to extract.
Adobe's Acrobat help draws the same line. In the export dialog, choosing the option Export all Images extracts the images themselves. Without it, all pages of the PDF are saved in the chosen image format. Acrobat can export raster images but not vector objects, and it has an option to exclude images smaller than a chosen size.
Why the files look strange
The PDF format describes pictures as image objects, and the specification gives them several optional parts. ISO 32000-2 lists a colour space for the image (DeviceGray, DeviceRGB or DeviceCMYK are built in), a filter such as DCTDecode for JPEG data, and separate objects used for transparency: an image mask, set with the ImageMask flag, and a soft mask, set with the SMask entry, which is described as a subsidiary image object. This explains the common oddities.
- Two files for one picture. Transparency is stored as a separate image. The poppler tool pdfimages states that transparency is represented in PDF using a separate image for the image and the mask or soft mask, and that the mask always immediately follows the image in its list. Its list shows a type for each entry: image (an opaque image), mask (a monochrome mask), smask (a soft mask) and stencil (a monochrome mask used for painting a colour or pattern). A black and white shape that looks like a silhouette is usually one of these masks.
- Pictures in colours that look wrong. CMYK is the colour model used for print work. pdfimages writes CMYK images as TIFF files with its -all option. Screens work in RGB, so converting a CMYK picture to RGB can shift some colours.
- Many small pieces. Some programs cut one large picture into strips or tiles, or paint a page from many small objects. Each is a separate image object, so an extractor returns many files. If the pieces are not wanted, convert the page instead.
- Different formats than expected. The PDF may hold JPEG, JPEG 2000, JBIG2 or CCITT data. pdfimages with -all writes these in their native format. Its default output is PBM or PPM instead.
- No picture at all. Drawings, charts made of lines and text are not image objects.
Scanned PDFs
Scanner and phone apps usually store each page of a scan as one large image, often with an invisible text layer on top if the PDF is searchable. Extracting gives the page images at the scanned size, which can be larger than what a 300 dpi render of the page would give. For a scan, extraction is often the way to get the clean page pictures. If the text layer is the goal, see how to extract text from a scanned document.
Extract the pictures with DocPro
- Open Extract images from PDF and choose the PDF.
- In Pages, leave the field blank for all pages, or type a range such as 1-3, 5, 8-.
- Set Ignore images smaller than (px). The default is 32. It skips tiny icons, bullets and line fragments. Use 0 to keep everything.
- Press Find images. Each picture appears as a thumbnail with its page and pixel size, for example p3 and the width and height.
- Tick the pictures you want, or use All and None.
- Choose Format: PNG or JPEG. PNG keeps transparency. JPEG has no transparency, and transparent areas show on white.
- Press Save selected. Several images download as a ZIP.
A picture repeated on many pages, such as a logo, is listed once.
What Extract images does to the file
DocPro reads the image objects with the pdf.js library in the browser. It decodes each picture to pixels, then saves those pixels as a new PNG or JPEG. This has consequences worth knowing:
- The pixel size is kept. The width and height shown on the thumbnail are the size of the file you save.
- The file is not a byte-for-byte copy. A JPEG inside the PDF is decoded and, if you choose JPEG, encoded again at quality 92 percent, which causes a small extra loss. Choose PNG to avoid a second round of compression. PNG files are larger than the original JPEG.
- Output is an ordinary RGB image from the browser, so a CMYK picture is converted to RGB for screens and does not keep a print colour profile.
- An image the viewer library cannot decode, or a format it does not support, is skipped. If the count looks too low, set the minimum size to 0 and check the page by eye.
- Masks are not offered as a separate kind of file, and how a particular transparent picture comes out depends on the PDF. Check the thumbnail, which sits on a chequered background so transparent areas show.
- The tool does not list vector drawings, text or page backgrounds made of drawn shapes.
If you must have the exact original JPEG, JPEG 2000 or CMYK data, use a tool that writes native data. The command line tool pdfimages from poppler does this with -all. Acrobat's Export all Images is another option. Both run outside the browser.
Save whole pages as pictures instead
When you want the page as a picture, use PDF to images or PDF to JPG, which pre-selects JPEG.
- Choose the PDF.
- Set Format to PNG or JPEG. For JPEG, set JPEG quality between 30 and 100 percent.
- Pick Resolution: 72, 100, 150, 200 or 300 dpi. The line below shows the pixel size page 1 will have.
- Pick Pages, then Output: Separate files (ZIP) or One tall image.
- Press Convert to images.
150 dpi suits screens, and 300 dpi suits printing but makes much larger images. PNG is lossless and suits text. Very large pages are drawn slightly below the chosen resolution so that the browser does not run out of memory, and the tool tells you when that happens.
A note on rights
Pictures in a PDF usually belong to someone. Being able to extract a photo does not give permission to reuse it. The US Copyright Office advises that in cases of doubt permission be obtained, and says you cannot claim copyright to another's work, no matter how much you change it, unless you have the owner's consent. Laws differ by country. Reuse images you made, images you have a licence for, or images the owner has cleared.
Not legal advice. For images you plan to publish or sell, ask the owner or a lawyer.
Frequently asked questions
How do I extract images from a PDF without losing quality?
Extract the embedded images instead of converting pages. Extraction keeps each picture at its stored pixel size. For a byte-exact copy, including the original JPEG data, use a tool that writes native data, such as pdfimages with -all or Acrobat's Export all Images. DocPro saves new PNG or JPEG files at the same pixel size, and PNG avoids extra compression.
Why does my extracted image have a black and white copy next to it?
The black and white file is usually a mask. PDFs store transparency as a separate image, and the mask follows the picture it belongs to. The picture file is the one you want.
Why is a picture saved from a page smaller than the one in the PDF?
A page snapshot has the resolution you pick, and a placed photo occupies only part of the page. At 300 dpi, a photo 8 cm wide is about 945 pixels wide, however large the original is. Extract the image to get the full size.
Can I extract the pictures from a scanned PDF?
Usually yes, because each scanned page is typically one image. The result is an image of the page, not the text. Use OCR if you need the text.
Is it legal to reuse pictures from a PDF?
Only with the right to do so. Extraction is a technical step and gives no permission. Check the licence or ask the owner.