How to Extract Images from a PDF Without Exporting Whole Pages

To extract images from a PDF, first decide whether you need the embedded picture files or a picture of each complete page. These are different operations. An image extractor can recover a photograph without its surrounding paragraph; a page-to-image conversion includes the whole visible page, including text, borders and captions.

For individual photographs, use an export option that explicitly extracts images, or an image-extraction library for a repeatable local workflow. For a diagram made from vector objects, you may need a page rendering or the original artwork instead. Keep the PDF as a reference so you can identify what each extracted file actually represents.

Choose extraction or page conversion

A PDF page can combine text, embedded raster images and drawing instructions. A visible figure is therefore not guaranteed to exist inside the document as one ready-made JPEG. It might be several image pieces, a vector chart, or a scanned image of the entire page.

If your actual goal is whole-page pictures, follow the PDF-to-JPG guide. Using an embedded-image extractor for that job can produce confusing fragments and omit text that was never part of an image resource.

Use extraction only for material you are authorized to reuse. Recovering an image file does not grant publication rights, remove attribution requirements or make a confidential illustration suitable for another audience.

Inspect one representative page first

Choose a page with a photograph, a caption and, if possible, a diagram. Note its page number and what you expect to recover. Testing a representative page helps separate a wrong export setting from a PDF whose internal structure simply differs from its appearance.

A scanned page may contain one large raster image with both the photograph and surrounding text already baked into it. Extracting that raster gives you the scan, not the original camera photograph. Conversely, a sharp chart may be drawn with vector instructions, leaving no corresponding image resource to extract.

Also note visible cropping and rotation. The stored image can contain material outside the portion displayed on the PDF page. An extracted resource may look wider or show a different orientation than the page presentation, so compare it with the source before deciding it is the wrong file.

Extract individual images in Acrobat

Adobe’s image-export documentation describes an important option: Export all images. Without it, the selected image format can be used for complete pages instead of individual embedded pictures. Interface labels and available export features depend on your Acrobat installation.

1. Open the PDF, select Convert, or choose Export a PDF from All tools.
2. Choose an image format appropriate to the result you need.
3. Enable Export all images. Inspect Settings, including any threshold excluding small images.
4. For a first diagnostic extraction, use No Limit where the documented size filter is available.
5. Export into a new empty folder and inspect the files alongside the source page.

Adobe states that this workflow exports raster images, not vector objects. A missing vector chart is therefore a limitation of the chosen operation, not proof that the page is damaged. Do not disable sandbox protection or other security features to work around a filename inconvenience.

Use a new folder for each attempt. Otherwise an older page rendering can sit beside a newer extracted photograph, making the export appear to have produced inconsistent results. Record the export format and any exclusion threshold if you need to repeat the task later.

A local Python method for repeatable extraction

The pypdf image-extraction guide exposes images through each page’s images collection and notes optional dependencies. For a Python environment you control, install the image support with python -m pip install "pypdf[image]", subject to your normal software approval process.

The following example saves decoded page images as PNG files. It deliberately generates filenames rather than trusting embedded names, which can be repeated or contain unsuitable path characters. It normalizes the output format; it does not promise byte-for-byte copies of the original compressed resources.

from pathlib import Path
from pypdf import PdfReader

reader = PdfReader("input.pdf")
if reader.is_encrypted:
    raise ValueError("Use an authorized readable source copy.")

out = Path("extracted-images")
out.mkdir(exist_ok=False)
for page_number, page in enumerate(reader.pages, 1):
    for image_number, item in enumerate(page.images, 1):
        image = item.image
        if image.mode not in ("1", "L", "LA", "P", "RGB", "RGBA"):
            image = image.convert("RGB")
        name = f"page-{page_number:04d}-image-{image_number:03d}.png"
        image.save(out / name, format="PNG")

Run it with the PDF saved as input.pdf in your working folder. The output directory must not already exist; this avoids silently overwriting a previous extraction. For another attempt, rename the old output folder or choose a different new directory. The generated page numbers refer to physical page positions, not printed labels such as “iv.”

We checked the extraction loop locally on October 17, 2026 with pypdf 6.10.0 and a synthetic two-page fixture. Each page contained a 160-by-90-pixel bitmap, text and a vector rectangle. The loop produced two PNG files at 160 by 90 pixels. That limited check demonstrates this example on that fixture; it is not a claim that every PDF or image type has been tested.

For a large or unfamiliar document, try a copy containing a small representative page range first. Image decoding may use substantial resources, and malformed or unsupported content can stop processing. A partially populated output folder is not evidence that every page completed.

Understand common extraction surprises

The same logo appears many times

A resource may be used on several pages, and a page-based loop can produce repeated results. Retain the page-based names until you have checked placement and purpose. If you later deduplicate files, compare their actual content rather than assuming matching dimensions mean the images are identical.

The export contains tiny files or unexpected fragments

Decorative elements and image pieces can be separate resources. Start without a size filter so you can understand what exists. Then choose exclusions based on the material you need, rather than setting an arbitrary threshold that might discard a small but meaningful icon or stamp.

A transparent image looks wrong

Some PDFs use a separate mask to describe transparency. The PyMuPDF image recipes discuss reconstructing images that use masks. If your output has an unexpected background or missing transparent areas, compare the decoding method with that structure before editing pixels manually.

A chart is missing entirely

The chart may consist of vector drawing commands rather than a raster image. PyMuPDF’s project documentation explicitly distinguishes these objects. Obtain the original chart or choose a rendering workflow when a faithful visual copy is sufficient. Neither method recovers an underlying spreadsheet automatically.

The extracted image is blurrier than expected

Check its actual pixel dimensions. Extraction cannot invent detail that was discarded when the PDF was created. A small image can look acceptable at a small display size while becoming visibly soft when enlarged. If you need a larger reproduction, request the original image rather than repeatedly resaving the extracted copy.

Review the files before reuse

Make a simple inventory with the output filename, source page and intended use. Open each image you plan to reuse, not merely its thumbnail. Compare edges, orientation, transparency and important labels with the PDF. Keep a separate note for figures that were rendered rather than extracted.

Check captions and attribution separately. Text surrounding an image often remains PDF text rather than becoming part of the exported picture. A photograph without its credit line may be technically usable but unsuitable for your publication workflow. Preserve the information needed to assess rights and identify its original context.

For a confidential source, inspect the full extracted resource for portions hidden by the PDF’s visible crop. Do not assume that only the displayed region was retained. If the source requires sensitive-content removal, consult the PDF redaction guide and use an approved process before sharing a derivative.

Example: recover figures from a project report

Suppose a ten-page report has three photographs, a vector chart and a scanned appendix. Make an inventory before exporting: photograph A on page two, photographs B and C on page five, the chart on page seven and the scan on page ten. These are illustrative contents, not measured results from a customer document.

After extraction, do not judge success by whether five files appeared. One photograph could be split into pieces, a recurring logo could create additional files, and the chart might not appear as a raster resource. Match each required figure to a source page and an output file. Mark the vector chart as needing a different method rather than calling the whole extraction unsuccessful.

For the scanned appendix, decide whether a complete page image is useful. If the task is to reuse only a photograph printed within that scan, an authorized crop can isolate its appearance, but it remains a crop of a scan. Name it accordingly in your inventory so a later editor does not mistake it for the original photographic file.

Finish by reviewing the three selected photographs at their intended display sizes and carrying their captions and credits into the handoff notes. Send only the approved outputs, with a reference to the source revision. This gives the recipient a manageable figure package instead of an unexplained folder of every raster resource the extractor encountered.

Questions that help choose the next step

Is extraction the same as taking a screenshot?

No. A screenshot records a rendered view. Resource extraction retrieves image content from the PDF’s structure. Either can be appropriate, but they answer different needs and should not be labeled interchangeably.

Does OCR recover the original photograph?

No. OCR concerns recognizing text in image-based content. It does not recreate a separate high-resolution photograph from a scanned page. Request the source image when the extracted scan is insufficient.

Should I export JPEG or PNG?

Choose according to the output’s purpose, then inspect it. PNG is useful for the normalized example above; it is not a guarantee of smaller files or identical appearance for every source color space. Keep original resources when your workflow requires their original encoding and use a method designed for that requirement.

Official references

• Adobe: Export images from PDFs
• pypdf: Extract images
• PyMuPDF: Image recipes
• PyMuPDF project documentation

Sources checked October 17, 2026. Featured photo by Tanja Tepavac, via Unsplash, under the Unsplash License. This document photograph is illustrative, not a software screenshot.