Extract individual field from table image to excel with OCR

Question

I have scanned images which have tables as shown in this image:

I am trying to extract each box separately and perform OCR but when I try to detect horizont

Answer 1

This is function, which uses tesseract-ocr for layout detection. You can try with different RIL levels and PSM. For more details have a look here: https://github.com/sirfz/tesserocr

import os
import platform
from typing import List, Tuple

from tesserocr import PyTessBaseAPI, iterate_level, RIL

system = platform.system()
if system == 'Linux':
    tessdata_folder_default = ''
elif system == 'Windows':
    tessdata_folder_default = r'C:\Program Files (x86)\Tesseract-OCR\tessdata'
else:
    raise NotImplementedError

# this tesseract specific env variable takes precedence for tessdata folder location selection
# especially important for windows, as we don't know if we're running 32 or 64bit tesseract
tessdata_folder = os.getenv('TESSDATA_PREFIX', tessdata_folder_default)


def get_layout_boxes(input_image,  # PIL image object
                     level: RIL,
                     include_text: bool,
                     include_boxes: bool,
                     language: str,
                     psm: int,
                     tessdata_path='') -> List[Tuple]:
    """
    Get image components coordinates. It will return also text if include_text is True.
    :param input_image: input PIL image
    :param level: page iterator level, please see "RIL" enum
    :param include_text: if True return boxes texts
    :param include_boxes: if True return boxes coordinates
    :param language: language for OCR
    :param psm: page segmentation mode, by default it is PSM.AUTO which is 3
    :param tessdata_path: the path to the tessdata folder
    :return: list of tuples: [((x1, y1, x2, y2), text)), ...]
    """
    assert any((include_text, include_boxes)), (
        'Both include_text and include_boxes can not be False.')

    if not tessdata_path:
        tessdata_path = tessdata_folder

    try:
        with PyTessBaseAPI(path=tessdata_path, lang=language) as api:
            api.SetImage(input_image)

            api.SetPageSegMode(psm)
            api.Recognize()
            page_iterator = api.GetIterator()
            data = []
            for pi in iterate_level(page_iterator, level):
                bounding_box = pi.BoundingBox(level)
                if bounding_box is not None:
                    text = pi.GetUTF8Text(level) if include_text else None
                    box = bounding_box if include_boxes else None
                    data.append((box, text))
            return data
    except RuntimeError:
        print('Please specify correct path to tessdata.')