Extracting text contained inside a rect using quads

[andrescase]

Hi,

I was wondering if there is any way to extract any words contained within the area of a rect object or using coordinates in a PDF file.

Thank you in advance for your help,

[andrescase]

Was able to solve it by looking for words within coordinate interval.

[JorjMcKie]

I would do this:

• assume your rectangle is named rect
• make a list of words on the page: words = page.get_text("words").
• each item of the word list has 4 starting floats following by the word string: (x0, y0, x1, y1, "word",...) followed by some more info

for word in words:
    bbox = fitz.Rect(word[:4])  # make bbox from subtuple
    if bbox in rect:
        print("word '%s' in rectangle" % w[4])

[JorjMcKie]

If you need to check if a word is contained in a quad, and not necessarily in a rectangle, things are currently more complicated.

• First you have to map the quad such that its sides are parallel to the x- resp. y-axis. If the quad is more general, i.e. has no corners with 90° angles, then there is no solution at all, currently. But you can check whether a quad is congruent to a rectangle via quad.is_rectangular.
• There are ways to compute a matrix such that the line from quad.ul to quad.ur lies on the x-axis. This matrix (let us call it mat) thus converts the complete quad to a traditional rectangle.
• Then check whether bbox * mat is in the rectangle of the converted quad.

mat = fitz.planish_line(quad.ul, quad.ur)  # compute the conversion matrix
rect = (quad * mat).rect  # rect of converted quad

# ...
for word in words:
    bbox = fitz.Rect(word[:4])  # make bbox from subtuple
    if bbox * mat in rect:
        print("word '%s' in quad" % w[4])

A future version of PyMuPDF will allow to conveniently check, whether a point is inside a quad. Then you would only need to check whether all bbox corners are inside the quad.

[JorjMcKie]

Oops - sorry: I forgot I already have that said function included 🙄 - you can do this:

for word in words:
    bbox = fitz.Rect(word[:4])  # make bbox from subtuple
    if bbox in quad:
        print("word '%s' in quad" % w[4])

[JorjMcKie]

All of the following checks are possible:

• point in quad
• rect in quad
• quad in quad

[JorjMcKie]

I have just updated the documentation: you will find my previous comments in the section for class Quad.