extract_img_pdf.py

#==================================================================================#
#                    Find Contours from Image and Convert into PDF                 #
#==================================================================================#
import cv2, os
import numpy as np
from imutils.perspective import four_point_transform
from PIL import Image 
from unstructured.partition.pdf import partition_pdf
import json, base64, io
from flask import Flask, render_template, flash, redirect, url_for
from dotenv import load_dotenv
import pytesseract

load_dotenv()

app = Flask(__name__)
app.secret_key = os.getenv("SECRET_KEY")
pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"
poppler_path=r"C:\poppler-23.11.0\Library\bin"

count = 0

OUTPUT_FOLDER = "OUTPUTS"
# os.makedirs(OUTPUT_FOLDER, exist_ok=True)
IMAGE_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "SCANNED_IMAGE")
# os.makedirs(IMAGE_FOLDER_PATH, exist_ok=True)

PDF_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "SCANNED_PDF")
# os.makedirs(PDF_FOLDER_PATH, exist_ok=True)
JSON_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "EXTRACTED_JSON")

for path in [OUTPUT_FOLDER, IMAGE_FOLDER_PATH, PDF_FOLDER_PATH, JSON_FOLDER_PATH]:
    os.makedirs(path, exist_ok=True)

# --- FUNCTION: Detect document contour ---
def detect_document_contour(image):
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    blur = cv2.GaussianBlur(gray, (5, 5), 0)
    _, thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

    contours, _ = cv2.findContours(thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
    contours = sorted(contours, key=cv2.contourArea, reverse=True)

    for contour in contours:
        area = cv2.contourArea(contour)
        if area > 1000:
            peri = cv2.arcLength(contour, True)
            approx = cv2.approxPolyDP(contour, 0.02 * peri, True)
            if len(approx) == 4:
                return approx
    return None

# --- FUNCTION: Extract images from saved PDF ---
def extract_images_from_pdf(pdf_path, output_json_path):
    elements = partition_pdf(
        filename=pdf_path,
        
        strategy="hi_res",
        extract_image_block_types=["Image"],  # or ["Image", "Table"]
        extract_image_block_to_payload=True,  # Set to True to get base64 in output
    )
    with open(output_json_path, "w") as f:
        json.dump([element.to_dict() for element in elements], f, indent=4)

    # Display extracted images
    with open(output_json_path, 'r') as file:
        file_elements = json.load(file)
        
    for i, element in enumerate(file_elements):
        if "image_base64" in element["metadata"]:
            image_data = base64.b64decode(element["metadata"]["image_base64"])
            image = Image.open(io.BytesIO(image_data))
            image.show(title=f"Extracted Image {i+1}")

# --- Route: Home Page ---
@app.route("/")
def index():
    return render_template("index.html")

# --- Route: Scan Document ---
@app.route("/scan")
def scan_document():
    global count

    cap = cv2.VideoCapture(0 + cv2.CAP_DSHOW)
    cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
    cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)

    scale = 0.5
    contour = None

    while True:
        ret, frame = cap.read()
        if not ret:
            flash("Camera Error!", "error")
            break

        frame = cv2.rotate(frame, cv2.ROTATE_180)
        display = frame.copy()
        contour = detect_document_contour(display)

        if contour is not None:
            cv2.drawContours(display, [contour], -1, (0, 255, 0), 3)

        resized = cv2.resize(display, (int(scale * display.shape[1]), int(scale * display.shape[0])))
        cv2.imshow("📷 Scan Document - Press 's' to Save, ESC to Exit", resized)

        key = cv2.waitKey(1) & 0xFF
        if key == 27:  # ESC
            break
        elif key == ord('s') and contour is not None:
            warped = four_point_transform(frame, contour.reshape(4, 2))
            image_path = os.path.join(IMAGE_FOLDER_PATH, f"scanned_colored_{count}.jpg")
            pdf_path = os.path.join(PDF_FOLDER_PATH, f"scanned_colored_{count}.pdf")
            json_path = os.path.join(JSON_FOLDER_PATH, f"scanned_{count}.json")

            cv2.imwrite(image_path, warped)
            img = Image.open(image_path).convert("RGB")
            img.save(pdf_path)
            extract_images_from_pdf(pdf_path, json_path)

            flash("✅ Document scanned and saved!", "success")
            count += 1
            break

    cap.release()
    cv2.destroyAllWindows()
    return redirect(url_for("index"))


# --- Run ---
if __name__ == "__main__":
     app.run(host="0.0.0.0", port=7860, debug=False)
# while True:
#     ret, frame = cap.read()
#     if not ret:
#         break

#     frame = cv2.rotate(frame, cv2.ROTATE_180)
#     display = frame.copy()

#     contour = detect_document_contour(display)
#     if contour is not None:
#         cv2.drawContours(display, [contour], -1, (0, 255, 0), 3)

#     cv2.imshow("Document Scanner", cv2.resize(display, (int(scale * display.shape[1]), int(scale * display.shape[0]))))

#     key = cv2.waitKey(1) & 0xFF

#     if key == 27:  # ESC to exit
#         break

#     elif key == ord('s') and contour is not None:
#         warped = four_point_transform(frame, contour.reshape(4, 2))
#         image_path = os.path.join(IMAGE_FOLDER_PATH, f"scanned_colored_{count}.jpg")
#         pdf_path = os.path.join(PDF_FOLDER_PATH,f"scanned_colored_{count}.pdf")
        
#         # Save the Image
#         cv2.imwrite(image_path, warped)
#         print(f"[INFO] Saved: {image_path}")
        
#         # Convert to PDF
#         img = Image.open(image_path)
#         img_rgb = img.convert("RGB")
#         img_rgb.save(pdf_path)
#         print(f"[INFO] Converted to PDF: {pdf_path}")
        
#         # Extract and show embedded images from PDF
#         print(f"[INFO] Extracting embedded images from PDF...")
#         # extract_images_from_pdf(pdf_path, JSON_FOLDER_PATH)
        
#         count += 1
# cap.release()
# cv2.destroyAllWindows()


''' Simple version Not a Flask APP '''
# import cv2, os, json, base64, io
# import numpy as np
# from imutils.perspective import four_point_transform
# from PIL import Image
# from unstructured.partition.pdf import partition_pdf
# import pytesseract

# # --- PATH CONFIGURATION ---
# pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"
# POPPLER_PATH = r"C:\poppler-23.11.0\Library\bin"

# OUTPUT_FOLDER = "OUTPUTS"
# IMAGE_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "SCANNED_IMAGE")
# PDF_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "SCANNED_PDF")
# JSON_OUTPUT_FOLDER = os.path.join(OUTPUT_FOLDER, "EXTRACTED_JSON")

# for path in [OUTPUT_FOLDER, IMAGE_FOLDER_PATH, PDF_FOLDER_PATH, JSON_OUTPUT_FOLDER]:
#     os.makedirs(path, exist_ok=True)

# # --- FUNCTION: Detect document contour ---
# def detect_document_contour(image):
#     gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
#     blur = cv2.GaussianBlur(gray, (5, 5), 0)
#     _, thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
#     contours, _ = cv2.findContours(thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
#     contours = sorted(contours, key=cv2.contourArea, reverse=True)

#     for contour in contours:
#         area = cv2.contourArea(contour)
#         if area > 1000:
#             peri = cv2.arcLength(contour, True)
#             approx = cv2.approxPolyDP(contour, 0.02 * peri, True)
#             if len(approx) == 4:
#                 return approx
#     return None

# # --- FUNCTION: Extract images from saved PDF ---
# def extract_images_from_pdf(pdf_path, output_json_path):
#     elements = partition_pdf(
#         filename=pdf_path,
#         poppler_path=POPPLER_PATH,
#         strategy="hi_res",
#         extract_image_block_types=["Image"],
#         extract_image_block_to_payload=True,
#     )
    
#     with open(output_json_path, "w") as f:
#         json.dump([element.to_dict() for element in elements], f, indent=4)

#     # Display extracted images
#     with open(output_json_path, 'r') as file:
#         file_elements = json.load(file)

#     for i, element in enumerate(file_elements):
#         if "image_base64" in element["metadata"]:
#             image_data = base64.b64decode(element["metadata"]["image_base64"])
#             image = Image.open(io.BytesIO(image_data))
#             image.show(title=f"Extracted Image {i+1}")

# # --- WEBCAM SCANNER START ---
# # cap = cv2.VideoCapture(0 + cv2.CAP_DSHOW)
# cap = cv2.VideoCapture("http://100.71.6.36:8080/video")
# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)

# scale = 0.5
# count = 0

# while True:
#     ret, frame = cap.read()
#     if not ret:
#         break

#     frame = cv2.rotate(frame, cv2.ROTATE_180)
#     display = frame.copy()

#     contour = detect_document_contour(display)
#     if contour is not None:
#         cv2.drawContours(display, [contour], -1, (0, 255, 0), 3)

#     cv2.imshow("Document Scanner", cv2.resize(display, (int(scale * display.shape[1]), int(scale * display.shape[0]))))

#     key = cv2.waitKey(1) & 0xFF

#     if key == 27:  # ESC to exit
#         break

#     elif key == ord('s') and contour is not None:
#         warped = four_point_transform(frame, contour.reshape(4, 2))
        
#         image_path = os.path.join(IMAGE_FOLDER_PATH, f"scanned_colored_{count}.jpg")
#         pdf_path = os.path.join(PDF_FOLDER_PATH, f"scanned_colored_{count}.pdf")
#         json_path = os.path.join(JSON_OUTPUT_FOLDER, f"embedded_images_{count}.json")

#         # Save Image
#         cv2.imwrite(image_path, warped)
#         print(f"[INFO] Saved image: {image_path}")

#         # Convert to PDF
#         img = Image.open(image_path)
#         img_rgb = img.convert("RGB")
#         img_rgb.save(pdf_path)
#         print(f"[INFO] Converted to PDF: {pdf_path}")

#         # Extract and show embedded images from PDF
#         print(f"[INFO] Extracting embedded images from PDF...")
#         extract_images_from_pdf(pdf_path, json_path)

#         count += 1

# cap.release()
# cv2.destroyAllWindows()



'''
#==================================================================================#
#                               Extract Images from PDF                            #
#==================================================================================#
from unstructured.partition.pdf import partition_pdf
import pytesseract
pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"

elements = partition_pdf(
    filename=r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images\page1.pdf",
    poppler_path=r"C:\poppler-23.11.0\Library\bin",
    strategy="hi_res",
    extract_image_block_types=["Image"],  # or ["Image", "Table"]
    extract_image_block_to_payload=True,  # Set to True to get base64 in output
)

import json, base64, io, os
from PIL import Image

# Save JSON output
os.makedirs("output", exist_ok=True)
with open("output/embedded-images-tables.json", "w") as f:
    json.dump([element.to_dict() for element in elements], f, indent=4)


def get_image_block_types(input_json_file_path: str):
    with open(input_json_file_path, 'r') as file:
        file_elements = json.load(file)

    for element in file_elements:
        if "image_base64" in element["metadata"]:
            image_data = base64.b64decode(element["metadata"]["image_base64"])
            image = Image.open(io.BytesIO(image_data))
            image.show()

# Example usage:
get_image_block_types("output/embedded-images-tables.json")'''

# from unstructured_client import UnstructuredClient
# from unstructured_client.models import operations, shared
# from unstructured.staging.base import elements_from_dicts, elements_to_json

# import os
# import base64
# from PIL import Image
# import io

# if __name__ == "__main__":
#     client = UnstructuredClient(
#         api_key_auth=os.getenv("UNSTRUCTURED_API_KEY")
#     )

#     # Path to your PDF file
#     local_input_filepath = "your-pdf-file.pdf"
#     local_output_filepath = "output.json"

#     with open(local_input_filepath, "rb") as f:
#         files = shared.Files(
#             content=f.read(),
#             file_name=local_input_filepath
#         )

#     request = operations.PartitionRequest(
#         shared.PartitionParameters(
#             files=files,
#             split_pdf_page=True,
#             split_pdf_allow_failed=True,
#             split_pdf_concurrency_level=15,
#             # Extract Base64-encoded images and tables
#             extract_image_block_types=["Image", "Table"]
#         )
#     )

#     try:
#         result = client.general.partition(request=request)

#         for element in result.elements:
#             if "image_base64" in element["metadata"]:
#                 # Decode and display the image
#                 image_data = base64.b64decode(element["metadata"]["image_base64"])
#                 image = Image.open(io.BytesIO(image_data))
#                 image.show()  # This will open the image
        
#         # Save results as JSON
#         dict_elements = elements_from_dicts(element_dicts=result.elements)
#         elements_to_json(
#             elements=dict_elements,
#             indent=2,
#             filename=local_output_filepath
#         )
#     except Exception as e:
#         print(e)

# -------------------------------------------------------------------------------------- #

# # STEP 1
# # import libraries
# import fitz  # PyMuPDF
# import io
# from PIL import Image

# # STEP 2
# # file path you want to extract images from
# file = r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images/page1_orig.pdf"

# # open the file
# pdf_file = fitz.open(file)

# # STEP 3
# # iterate over PDF pages
# for page_index in range(len(pdf_file)):

#     # get the page itself
#     page = pdf_file.load_page(page_index)  # load the page
#     image_list = page.get_images(full=True)  # get images on the page

#     # printing number of images found in this page
#     if image_list:
#         print(f"[+] Found a total of {len(image_list)} images on page {page_index}")
#     else:
#         print("[!] No images found on page", page_index)
    
#     for image_index, img in enumerate(image_list, start=1):
#         # get the XREF of the image
#         xref = img[0]

#         # extract the image bytes
#         base_image = pdf_file.extract_image(xref)
#         image_bytes = base_image["image"]

#         # get the image extension
#         image_ext = base_image["ext"]

#         # save the image
#         image_name = f"image{page_index+1}_{image_index}.{image_ext}"
#         with open(image_name, "wb") as image_file:
#             image_file.write(image_bytes)
#             print(f"[+] Image saved as {image_name}")
            
# -------------------------------------------------------------------------------------- #

# from pdf2image import convert_from_path
# import numpy as np
# import cv2

# def extract_grid_cells_from_pdf(pdf_path, prefix="sub"):
#     # Convert PDF's first page to image
#     pages = convert_from_path(
#         pdf_path,
#         dpi=300,
#         poppler_path=r"C:\poppler-23.11.0\Library\bin"  
#     )
#     pil = pages[0]
#     img = np.array(pil)[:, :, ::-1]  # RGB→BGR

#     gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
#     _, thresh = cv2.threshold(gray, 240, 255, cv2.THRESH_BINARY_INV)
#     kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5,5))
#     dil = cv2.dilate(thresh, kernel, iterations=2)

#     cnts, _ = cv2.findContours(dil, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
#     cells = [cv2.boundingRect(c) for c in cnts if cv2.contourArea(c) > 1000]
#     cells = sorted(cells, key=lambda r: (r[1], r[0]))

#     for i, (x, y, w, h) in enumerate(cells):
#         crop = img[y:y+h, x:x+w]
#         cv2.imwrite(f"{prefix}_{i:02d}.png", crop)
#         print("Saved", f"{prefix}_{i:02d}.png")

# if __name__ == "__main__":
#     extract_grid_cells_from_pdf(
#       r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images\page1_orig.pdf"
#     )


# import cv2
# import layoutparser as lp
# from pdf2image import convert_from_path
# from reportlab.pdfgen import canvas
# from reportlab.lib.pagesizes import letter
# import numpy as np
# import tempfile
# import os

# # 1️⃣ Setup LayoutParser model
# model = lp.Detectron2LayoutModel(
#     "lp://PrimaLayout/PrimaLayout/mask_rcnn_R_50_FPN_3x/config",
#     label_map={0: "Text", 1: "Title", 2: "List", 3: "Table", 4: "Figure"}
# )

# # 2️⃣ Utility to crop and save a layout region
# def crop_and_save(img, block, out_dir, idx):
#     x1, y1, x2, y2 = map(int, block.block.x_1_y_2_x_2_y_2)
#     cropped = img[y1:y2, x1:x2]
#     path = os.path.join(out_dir, f"crop_{idx}.png")
#     cv2.imwrite(path, cropped)
#     return path

# # 3️⃣ Convert cropped images into multi-page PDF
# def imgs_to_pdf(img_paths, output_pdf):
#     c = canvas.Canvas(output_pdf, pagesize=letter)
#     w, h = letter
#     for img in img_paths:
#         c.drawImage(img, 0, 0, width=w, height=h)
#         c.showPage()
#     c.save()

# # 4️⃣ If user input is a PDF or image folder
# def process_document(pdf_path, output_pdf):
#     imgs = convert_from_path(pdf_path)
#     cropped_paths = []
#     with tempfile.TemporaryDirectory() as tmp:
#         for page_idx, pil_im in enumerate(imgs):
#             img = cv2.cvtColor(np.array(pil_im), cv2.COLOR_RGB2BGR)
#             layout = model.detect(img)
#             for idx, block in enumerate(layout):
#                 path = crop_and_save(img, block, tmp, f"{page_idx}_{idx}")
#                 cropped_paths.append(path)
#         imgs_to_pdf(cropped_paths, output_pdf)

# # 5️⃣ Real-time camera/video feed
# def process_video(output_pdf, src=0, frame_limit=100):
#     cap = cv2.VideoCapture(src)
#     idx = 0
#     cropped_paths = []
#     with tempfile.TemporaryDirectory() as tmp:
#         while idx < frame_limit:
#             ret, img = cap.read()
#             if not ret:
#                 break
#             layout = model.detect(img)
#             for i, block in enumerate(layout):
#                 path = crop_and_save(img, block, tmp, f"{idx}_{i}")
#                 cropped_paths.append(path)
#             idx += 1
#         cap.release()
#         imgs_to_pdf(cropped_paths, output_pdf)

# if __name__ == "__main__":
#     import argparse

#     ap = argparse.ArgumentParser()
#     ap.add_argument("--input", required=True,
#                     help="path to PDF or 'cam' for camera")
#     ap.add_argument("--output", required=True, help="output PDF path")
#     ap.add_argument("--frames", type=int, default=50,
#                     help="frames to scan if using camera")
#     args = ap.parse_args()

#     if args.input.lower().endswith(".pdf"):
#         process_document(args.input, args.output)
#     elif args.input.lower() == "cam":
#         process_video(args.output, src=0, frame_limit=args.frames)
#     else:
#         print("Unsupported input. Use PDF path or 'cam'.")

# import cv2
# from PIL import Image
# import numpy as np

# def get_contours(frame):
#     gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
#     # Threshold to binary
#     _, thresh = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY_INV)
#     # Find contours
#     contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
#     return contours

# def extract_regions(frame, contours):
#     rois = []
#     for cnt in contours:
#         x, y, w, h = cv2.boundingRect(cnt)
#         if w*h < 1000:  # skip small noise
#             continue
#         roi = frame[y:y+h, x:x+w]
#         rois.append(roi)
#     return rois

# def save_rois_as_pdf(rois, output_path):
#     pil_imgs = []
#     for roi in rois:
#         rgb = cv2.cvtColor(roi, cv2.COLOR_BGR2RGB)
#         pil = Image.fromarray(rgb)
#         pil_imgs.append(pil)
#     if pil_imgs:
#         pil_imgs[0].save(output_path, save_all=True, append_images=pil_imgs[1:])
#         print(f"Saved {len(pil_imgs)} regions to {output_path}")

# def main():
#     cap = cv2.VideoCapture(0)
#     all_rois = []
#     print("Press 'c' to capture and extract; 'q' to quit.")

#     while True:
#         ret, frame = cap.read()
#         if not ret:
#             break
#         cv2.imshow("Live Feed", frame)

#         key = cv2.waitKey(1) & 0xFF
#         if key == ord('c'):
#             contours = get_contours(frame)
#             rois = extract_regions(frame, contours)
#             all_rois.extend(rois)
#             print(f"Captured {len(rois)} regions.")
#         elif key == ord('q'):
#             break

#     cap.release()
#     cv2.destroyAllWindows()

#     if all_rois:
#         save_rois_as_pdf(all_rois, "output_contours.pdf")
#     else:
#         print("No regions captured.")

# if __name__ == "__main__":
#     main()

# import cv2
# from PIL import Image
# import numpy as np

# def get_edge_contours(frame, low=50, high=150):
#     gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
#     blurred = cv2.GaussianBlur(gray, (5, 5), 1.0)
#     edges = cv2.Canny(blurred, low, high)
#     contours, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
#     return contours, edges

# def extract_edge_rois(frame, contours, min_area=1000):
#     rois = []
#     for cnt in contours:
#         x, y, w, h = cv2.boundingRect(cnt)
#         if w * h < min_area:
#             continue
#         roi = frame[y:y+h, x:x+w]
#         rois.append(roi)
#     return rois

# def save_rois_as_pdf(rois, output_path):
#     pil_imgs = []
#     for roi in rois:
#         rgb = cv2.cvtColor(roi, cv2.COLOR_BGR2RGB)
#         pil_imgs.append(Image.fromarray(rgb))
#     if pil_imgs:
#         pil_imgs[0].save(output_path, save_all=True, append_images=pil_imgs[1:])
#         print(f"✅ Saved {len(pil_imgs)} edge-region(s) to {output_path}")
#     else:
#         print("⚠️ No edge-based regions detected—PDF not created.")

# def main():
#     cap = cv2.VideoCapture(0)
#     all_rois = []
#     print("Press ‘c’ to capture current edge regions, ‘q’ to quit.")

#     while True:
#         ret, frame = cap.read()
#         if not ret:
#             break

#         contours, edges = get_edge_contours(frame)
#         cv2.imshow("Live Feed", frame)
#         cv2.imshow("Edges", edges)

#         key = cv2.waitKey(1) & 0xFF
#         if key == ord('c'):
#             rois = extract_edge_rois(frame, contours)
#             all_rois.extend(rois)
#             print(f"🔄 Captured {len(rois)} edge-region(s). Total: {len(all_rois)}")
#         elif key == ord('q'):
#             break

#     cap.release()
#     cv2.destroyAllWindows()

#     if all_rois:
#         save_rois_as_pdf(all_rois, "edge_contours.pdf")
#     else:
#         print("❌ No regions captured.")

# if __name__ == "__main__":
#     main()