Python & OpenCV

Question

Is there any Native Supports for Grabbing Images from PDFs or Create some sort of Object in Python that can contain the Images from a pdf that then can be access via OpenCV? I've looked at some scripts to dump the Images of a PDF into my directory but I'm aiming more at accessing the PDF and instead dumping the data from the PDF that is the image(s) into some sort of object I can access with OpenCV. My own exploration hasn't yielded any results so i figured I'd ask.

Added a Example of Using PyMuPDF based off example from @Ghilas BELHADJ

import fitz
import cv2
import numpy as np
from tkinter import Tk
from tkinter.filedialog import askopenfilename


class AccessPDF:

    def __init__(self):
        self.filepath = ""
        self.doc = None

    def openPDF(self):
        Tk().withdraw()
        self.filepath = askopenfilename()
        self.doc = fitz.open(self.filepath)

    def pixel2np(self,pix):
        im = np.frombuffer(pix.samples, dtype=np.uint8).reshape(pix.h, pix.w, pix.n)
        im = np.ascontiguousarray(im[..., [2, 1, 0]])  # rgb to bgr
        return im

    def displayKey(self):  
        pixobj = self.doc.getPagePixmap(0, alpha=False)
        im = self.pixel2np(pixobj)
        cv2.imwrite("testimg.png",im)
        cv2.imshow("Key" im)

Answer 1

Edit: I've made a modification in the code following the comment of @Dan Mašek

You can achieve this (load the PDF embedded images into OpenCV without writing intermediate objects on disk) using PyMuPDF and Numpy.

In this example, I'm using this pdf file.

import fitz
import cv2
import numpy as np


def pix2np(pix):
    im = np.frombuffer(pix.samples, dtype=np.uint8).reshape(pix.h, pix.w, pix.n)
    im = np.ascontiguousarray(im[..., [2, 1, 0]])  # rgb to bgr
    return im


doc = fitz.open('NGM_2018_Media_Kit.pdf')

# entire page
# pix = doc.getPagePixmap(0, alpha=False)

# first page , 5th image, xref element
pix = fitz.Pixmap(doc, doc.getPageImageList(0)[4][0])  
im = pix2np(pix)

cv2.putText(im, 'Azul fellawen', (100, 100),
            cv2.FONT_HERSHEY_SIMPLEX, 1.,
            (18, 156, 243), 2, cv2.LINE_AA)
cv2.imwrite('sample_0.png', im)

Answer 2

I've grabbed the images from an pdf containing images as well as text.

You can save the images using pix.writePNG() or just show it using cv2.imshow(), whichever suits you best.

import fitz    #pymupdf
from cv2 import cv2
import numpy as np

def pix2np(pix):
    im = np.frombuffer(pix.samples, dtype=np.uint8).reshape(pix.h, pix.w, pix.n)
    im = np.ascontiguousarray(im[..., [2, 1, 0]])  # rgb to bgr
    return im

def convertPdf(filename):  
    doc = fitz.open(filename)
    #count = 0
    for i in range(len(doc)):
        for img in doc.getPageImageList(i):
            xref = img[0]
            pix = fitz.Pixmap(doc, xref)

            #if pix.n < 5:       # this is GRAY or RGB
            # To save it to the disk
            #pix.writePNG(f"p{count}.png")

            im = pix2np(pix)
            cv2.imshow("image",im)
            cv2.waitKey(0)
            #count += 1
            pix = None

if __name__ == "__main__":
    filename = "sample.pdf"
    convertPdf(filename)