How to extract images while retaining Microsoft PowerPoint-like items layered on top of the images?

[aimanwafiq12]

Hi all,

Problem description

While trying to extract images from PDF documents, I am facing challenges on retaining any PowerPoint-like items or annotations layered on top of the images that I am trying to extract.

Sample image to be extracted

Below is the sample image from the PDF to be extracted (with annotations layered on top of the image):

Current Output

Below is the image that I am able to extract using my current code script:

Expected output

I need to extract out and save the images with the annotations (number and red box in red).

Code snippet

import fitz  # PyMuPDF
import json
import os

def extract_text_and_images(pdf_path, output_folder):
    doc = fitz.open(pdf_path)
    data = []

    # Create the output folder if it doesn't exist
    if not os.path.exists(output_folder):
        os.makedirs(output_folder)

    for page_num in range(len(doc)):
        page = doc.load_page(page_num)
        text = page.get_text("text")
        images = page.get_images(full=True)

        page_data = {
            "document_path": pdf_path,
            "text": text,
            "images": []
        }

        for img_index, img in enumerate(images):
            xref = img[0]
            base_image = doc.extract_image(xref)
            image_ext = base_image["ext"]
            if image_ext == "png":
                image_name = f"image_{page_num}_{img_index}.{image_ext}"
                image_path = os.path.join(output_folder, image_name)
                with open(image_path, 'wb') as img_file:
                    img_file.write(base_image["image"])
                page_data["images"].append({
                    "name": image_name,
                    "path": image_path
                })

        data.append(page_data)

    return data

def save_to_json(data, json_path):
    with open(json_path, 'w', encoding='utf-8') as json_file:
        json.dump(data, json_file, ensure_ascii=False, indent=4)

pdf_path = "sample-files/edited-for-ict-v2.pdf"
json_path = "extracted_data.json"
output_folder = "extracted_images"
extracted_data = extract_text_and_images(pdf_path, output_folder)
save_to_json(extracted_data, json_path)

print(f"Data has been saved to {json_path}")
print(f"Images have been saved to {output_folder}")

[JorjMcKie]

You again did not include the PDF in question. So I still don't understand the problem and I am forced to guess:
You have a PDF page show some image and some annotation(s) hovering above?

If true you must make a picture of the page (and not extract the image). First find the image's boundary box on the page page.get_image_rects(xref), then make a pixmap using the bbox as clip: pix = page.get_pixmap(clip=bbox).

[aimanwafiq12]

Thanks a lot! The solution you gave works great.

Here is my updated python code:

import fitz  # PyMuPDF
import json
import os

def extract_text_and_images_with_annotations(pdf_path, output_folder):
    doc = fitz.open(pdf_path)
    data = []

    # Create the output folder if it doesn't exist
    if not os.path.exists(output_folder):
        os.makedirs(output_folder)

    for page_num in range(len(doc)):
        page = doc.load_page(page_num)
        text = page.get_text("text")
        images = page.get_images(full=True)

        page_data = {
            "document_path": pdf_path,
            "text": text,
            "images": []
        }

        for img_index, img in enumerate(images):
            xref = img[0]
            bbox_list = page.get_image_rects(xref)
            
            # Check if there's a bounding box
            if bbox_list:
                # Take the first bounding box for the image (if there are multiple, you might need additional handling)
                bbox = bbox_list[0]

                # Generate a pixmap for the region including annotations (clipping to the image boundary)
                pix = page.get_pixmap(clip=bbox)

                # Save the pixmap as an image file only if the index ends with 1 or higher
                if img_index != 0:
                    image_name = f"annotated_image_{page_num}_{img_index}.png"
                    image_path = os.path.join(output_folder, image_name)
                    pix.save(image_path)
                    
                    # Append to page data
                    page_data["images"].append({
                        "name": image_name,
                        "path": image_path,
                        # "bbox": [bbox.x0, bbox.y0, bbox.x1, bbox.y1]
                    })

        data.append(page_data)

    return data

def save_to_json(data, json_path):
    with open(json_path, 'w', encoding='utf-8') as json_file:
        json.dump(data, json_file, ensure_ascii=False, indent=4)

# Usage
pdf_path = "sample-files/edited-for-ict-v2.pdf"
json_path = "extracted_data_v3.json"
output_folder = "extracted_images_v3"
extracted_data = extract_text_and_images_with_annotations(pdf_path, output_folder)
save_to_json(extracted_data, json_path)

print(f"Data has been saved to {json_path}")
print(f"Images with annotations have been saved to {output_folder}")

[jamie-lemon]

For what you need I think you could also use Page.get_annots() , https://pymupdf.readthedocs.io/en/latest/page.html#Page.annots and then do Annot.get_pixmap() , https://pymupdf.readthedocs.io/en/latest/annot.html#Annot.get_pixmap