Get redundant rows, columns in merged cell with color background

[dviettu134]

Description of the bug

When using find_tables, I got redundant rows, columns in merged cell with color background as a result of redundant detected lines.

How can I remove the redundant lines? Please refer to the attached PDF file.
Thank you.

sample.pdf

How to reproduce the bug

import fitz  # import PyMuPDF
from pymupdf import Page
if not hasattr(fitz.Page, "find_tables"):
    raise RuntimeError("This PyMuPDF version does not support the table feature")

def show_image(item, title=""):
    """Display a pixmap.

    Just to display Pixmap image of "item" - ignore the man behind the curtain.

    Args:
        item: any PyMuPDF object having a "get_pixmap" method.
        title: a string to be used as image title

    Generates an RGB Pixmap from item using a constant DPI and using matplotlib
    to show it inline of the notebook.
    """
    DPI = 150  # use this resolution
    import numpy as np
    import matplotlib.pyplot as plt

    # %matplotlib inline
    pix = item.get_pixmap(dpi=DPI)
    img = np.ndarray([pix.h, pix.w, 3], dtype=np.uint8, buffer=pix.samples_mv)
    plt.figure(dpi=DPI)  # set the figure's DPI
    plt.title(title)  # set title of image
    _ = plt.imshow(img, extent=(0, pix.w * 72 / DPI, pix.h * 72 / DPI, 0))

doc = fitz.open("sample.pdf")
page: Page = doc[0]

tabs = page.find_tables(edge_min_length=50)  # detect the tables
for i,tab in enumerate(tabs):  # iterate over all tables
    for cell in tab.header.cells:
        page.draw_rect(cell,color=fitz.pdfcolor["red"],width=0.3)
    page.draw_rect(tab.bbox,color=fitz.pdfcolor["green"])
    print(f"Table {i} column names: {tab.header.names}, external: {tab.header.external}")
    
show_image(page, f"Table & Header BBoxes")

PyMuPDF version

1.24.12

Operating system

Windows

Python version

3.9

[JorjMcKie]

In addition to the blue background in the header, the words inside have their own separate / additional background.
As the default detection strategy is to desperately look for any usable information, these word-based backgrounds are also taken into account (there is no color-based check to see if the word background color is any different from other background color).
To change this, specify strategy="lines_strict". This will ignore fill colors and only look at lines.