after crop page, then get_images()

[yonglee7015]

HI I just want to get_images in specify range in a page. So I crop the page, then use function get_images. But it still get all images in this page.

page.set_cropbox(fitz.Rect(100, 100, 550, 700))
image_list = page.get_images(full=True)

How to get images in specify range in a page?

[JorjMcKie]

page.get_images() may even list images that are not at all used by the page - not to mention what you are asking.

This is because that method looks at PDF definitions only and does not inspect the page's appearance source code.

To at least restrict the above list to images that are actually referenced by this page anywhere on the MediaBox, do a page.clean_contents() first.

To restrict that list to visible images (CropBox) do this:

imglist = page.get_images()
visibles = [item for item in imglist if page.get_image_rects(item[0])[0] in page.cropbox]

get_image_rects walks through the page's appearance instructions to determine each bbox of one image (given by its xref item[0]) on the page. In the above I am looking at the first such bbox only to see whether it is in the CropBox.
A more precise code would have to go over the potentially several such rectangles - and may even include images that are at least partially visible by checking non-empty rectangle intersections.

[yonglee7015]

Thank you so much. But it seems not as I want. I test it in my code

[yonglee7015]

Anyway I just want to extract images in a fixed position in pdf. based on rect can't get the correct image I need.

[JorjMcKie]

There must be some misconception. Can you let me have the PDF and the page number in question?

[yonglee7015]

Ok,just sent pdf to your email

[JorjMcKie]

As I suspected, you have been hiding major information items!
Your pages are rotated by 90°! As documented, every coordinates read or written however are (resp. must be) in unrotated coordinates.
The PDF CropBox itself always is in unrotated coordinates (in contrast to page.rect), so for setting this you need unrotated coords.
So if you extract image bboxes on a page, they will appear as if the page had not been rotated. This is also the coordinates you need for CropBox setting. If you do that and use this Rect(109.0, 241.76995849609375, 529.0, 550.219970703125) things will work fine.

[yonglee7015]

HI Jorj X. McKie, thanks for you help. Finally I found pdfplumber can solve my problem. Cos pdfplumber can export crop page to image directly. So I set a fixed crop bbox in the page,
Then it can always export the image I want.

!apt install imagemagick

import pdfplumber

pdf = pdfplumber.open('/content/015-23C110164.pdf')
currentpage = pdf.pages[0]
currentpage_crop = currentpage.crop((120,130,700,500))
image_obj = currentpage_crop.to_image(resolution=300)
image_obj.save('dddd.png', format="PNG")

[JorjMcKie]

Ah, it is only now that I understand what you actually wanted!

With the current solution you do not actually extract the underlying image of the PDF:
You are simply taking a "photo" of a certain page area.
That would have been even easier in PyMuPDF, if we had been communicating better:

pix = page.get_pixmap(dpi=300, clip=(120,130,700,500))
pix.save("dddd.png")

[yonglee7015]

Sorry for my poor English. I can't clearly express my meaning. Image rect is so dynamics,this way should be the best.

[JorjMcKie]

That's ok. BTW I am sure that PyMuPDF is faster here, too.

What brought me on the wrong track was that talk about changing the cropbox. But this was not at all what you wanted - you needed a subimage of the full page like a photo.

[yonglee7015]

Thanks. It is faster and simple. I need to read the pymupdf document carefully first.