Extract images from a PDF at original quality, or render the page?

Drop the PDF and press Extract pictures. It runs on your device and nothing is uploaded.

How a PDF stores a photo

A PDF page is a list of drawing instructions. Text is “these letters, in this font, here”. A photo is “take picture object 12 and show it in this rectangle”. The picture object holds its own pixels, at whatever size the author put in, and the page only says how big to show it.

That gap between stored size and shown size is the whole story. Our test page, a report page, shows a photo 17 cm wide. The object behind it is 3840 x 2400 px, which works out at about 570 pixels per inch on the page. A render of the page at 150 dpi gives that photo about 1000 px of width; at 300 dpi about 2000 px. Extraction gives all 3840.

What comes out

We ran Extract images on our test files on 11 October 2026:

The settings, as the panel names them:

• Format: As stored keeps JPEGs exactly as they are in the file and writes everything else as lossless PNG. All PNG decodes every picture to PNG; a JPEG then becomes a much larger PNG, but every file has the same format.
• Smallest: All, 50 px or 150 px. Pictures smaller than this on both sides are skipped, which keeps rules, bullets and icons out of your folder. 50 px is the default.
• Pages: all, or ranges such as 1-3, 7.

Files are named after the PDF, the page and the picture’s number on it (report-page3-2.jpg; in a PDF of 10 pages or more the page number gets a leading zero, report-page03-2.jpg, so the files sort in page order); several come as 1 zip. The result line says how many came out and how many are the stored JPEGs, for example “Took out 12 pictures at their own resolution, 9 of them as the original JPEG files”, or “all as the original JPEG files” when every one was. A picture used on many pages, such as a logo on every page, comes out once.

When extraction wins

• You want the photo itself. Product shots in a brochure, the pictures in a property listing, photos in a report. Extraction gives every pixel the author embedded, often several times what any render shows.
• You want no quality loss. A JPEG comes out byte for byte, so there is no second compression. A render makes a new JPG from the drawn page and compresses it again.
• The PDF is a scan. Each page is 1 stored picture, so extraction returns exactly what the scanner saved, at its own resolution, 1 file per page in page order.

When rendering wins

• You want the page as designed. Text, captions, several images together, a layout. That is a render: PDF to JPG or PDF to PNG.
• The “image” is vector art. Charts made in a spreadsheet, diagrams, most logos and icons are drawn as lines and shapes, so there is no picture object to extract. Render the page, cropped to the chart first if you only want the chart.
• The page changes how the picture looks. Extraction ignores where and how the page places a picture: a photo shown turned, flipped or cropped on the page comes out as it was stored, uncropped and unturned. If the page view is what you want, render it.
• The picture is see-through on the page. Some pictures get their transparency from a separate mask. Extraction does not apply that mask, so such a picture comes out on its own background. The page render shows it as intended; with PDF to PNG and a transparent background you keep the transparency too.

Check what a file holds first

Inspect lists a PDF’s images before you extract anything: how many there are, how much of the file they take, and the Largest image with its pixel size. If the largest image is a few hundred pixels, there is not much to gain over a 150 dpi render; if it is 4000 px wide, extraction is clearly the better way.

The honest part

• Odd colours can happen. Pictures in unusual colour spaces are decoded by the engine and written as PNG; CMYK JPEGs are decoded too, because many viewers show them wrongly. If an extracted picture looks off against the page, render the page at 300 dpi instead.
• Some programs store 1 photo in pieces. If you get many thin slices of the same picture, the PDF was made that way; render the page to get it whole.
• Nothing beats the source. A photo that was shrunk before it went into the PDF is small inside the PDF too; extraction cannot restore what the author threw away.
• Rights stay with the owner. Taking a photo out of someone’s brochure does not give you the right to reuse it.

For the full map of ways out of a PDF, see the pillar, PDF to images; for getting the most out of a render, PDF to JPG without losing quality.