How to Extract Text Tables Images from PDFs while maintaining the structures

from unstructured library opensource one when i tried a pdf that have background images design patterns and XObjects in it this library also consider those as a images and store the path. so how can we clean the pdfs and store only the images that are in the pdfs as a figure?

import os
from io import StringIO
from lxml import etree
import pandas as pd
from unstructured.partition.pdf import partition_pdf
import json

class PDFProcessor:
    def __init__(self):
        """Extract structured elements (text, images, tables, equations) from a PDF."""
        pass

    def process_pdf(self, path: str, max_characters=30000):
        try:
            pdf_elements = partition_pdf(
                filename=path,
                extract_images_in_pdf=True,
                strategy='hi_res',
                infer_table_structure=True,
                extract_image_block_types=["Image"],
                # extract_image_block_to_payload=True,
                max_character=max_characters,
            )
            chunks = [el.to_dict() for el in pdf_elements ]
            text_data = [el for el in chunks if el["type"] in ["NarrativeText", "Title", "ListItem", "Text", "FigureCaption","UncategorizedText"]]
            image_data = [el for el in chunks if el["type"] == "Image"]
            table_data = [el for el in chunks if el["type"] == "Table"]
            return {
                "text": text_data,
                "images": image_data,
                "tables": table_data
            }
        except Exception as e:
            print(f"[ERROR] Cannot process the document: {e}")
            return {
                "text": [],
                "images": [],
                "tables": [],
                "error": str(e)
            }