Pulls the actual embedded images out of a PDF, at the resolution they were stored, not screenshots of the page.
Screenshots of PDF pages cap quality at screen resolution and bake in scaling. The images embedded in a PDF are stored as discrete objects, often full-resolution photographs, and this tool pulls those objects directly through the reader engine's operator data, converting each to a downloadable PNG. The result is the best quality the file itself contains: if the source PDF embedded a 300-DPI photo, that is what downloads; if the PDF was built from low-resolution images, extraction faithfully returns those, because there is nothing better inside to give.
What counts as an 'image' here is the PDF's definition: embedded raster pictures. Vector graphics, text and page furniture are not images and are not extracted, a chart drawn with lines and text will not appear (the PDF to Image converter on this site renders whole pages when that is what you need). Some PDFs store one picture as several tiled pieces or with a separate transparency mask; tiles appear as separate downloads. Everything runs locally.
Why are some extracted images upside down or oddly colored? PDF image objects can store inverted or CMYK-ish data that viewers correct with page-level instructions. Rare, and the extraction is still the faithful stored data, flip or convert it in any editor.
The PDF has 40 pages of scans, is each page one image? Usually yes: scanned pages are stored as one full-page image each, and they will all extract, which is also a fine way to get page images at native resolution.
Is there a page/image limit? The tool caps the gallery at 80 images for browser sanity; huge documents are better handled in batches by splitting the PDF first (the Split PDF tool is on this site).