How to get image's location on page ?

[a22349]

use the page.getImageList just can know the width and height ,but not know the coordinates
use page.getText("blocks") can't get the image block neither if the page has image .
is there some api can get the image location ?

[JorjMcKie]

Yes! Please consult this method.

Use the following statements:

img_list = doc.getPageImageList(pno, full=True)  # important: use 'full' parameter
# for the required image item in this list, then do this:
rect = page.getImageBbox(item)  # where page = doc[pno]

[JorjMcKie]

Closed for lack of reaction.

[SuryaViswanath11]

Hey @JorjMcKie I am currently having a similar use case where I want to identify if a particular table has an image with it, however with the use of the getPageImageList() I will be able to get the coordinates. Is there a way I can check if there is an image next to text line so that I will be able to add that data into my excel

[JorjMcKie]

@SuryaViswanath11 - loop through the image list doc.get_page_images() (the new name of getPageImageList()) and store each image rectangle therein in a list (caution: not all images in the page list need actually be shown on the page - in that case the returned bbox is infinite).
Then check for each text line of interest, whether its bbox is close (enough) to one of the images in the stored list.

[rizwan29]

@JorjMcKie get_image_bbox method returns the Rect object which is a bounding box of the image but not the relative position of the image in the page. Is there a method which can help to extract image's location in the page?

[JorjMcKie]

not the relative position of the image in the page

Why do you say that? Of course it is!

[uttam-scai]

@rizwan29 were you able to get the position of the image with respect to the pdf page?
@JorjMcKie I am struggling to use the bbox output values in opencv (cv2.rectangle). It doesn't draw the box at the same location of the image.

[JorjMcKie]

@uttam-scai - have you confirmed that the bbox coordinates from PyMuPDF are correct?
You can do this by drawing that rectangle and inspecting it in a temp PDF:

page.draw_rect(bbox)
doc.save("temp.pdf")

If it works, then CV2 has a problem.
I am not using that package, so I cannot say what may be going wrong there.
What is CV2's page coordinate system: where is point (0, 0) - top-left or bottom-left?

[uttam-scai]

@JorjMcKie thank you very much for quick response. I have confirmed by using page.draw_rect command that it draws correctly. However it's failing in case of CV2. The reason of failure is the scale
in case of CV2:

• CV2's page coordinate system: top left is considered as (0,0), which is same as PyMuPDF as well.
  To draw the rectangle in cv2 and PyMuPDF, I use the following command -
• image = cv2.rectangle(img, (1, 1), (100, 100), color, thickness)
• page.draw_rect((1, 1, 100, 100))
  temp.pdf
  

However the pymupdf is covering more space than cv2. Is there any scale value that I can play around?
Please see the attached image and pdf that I received from the output.

Thank you.

[JorjMcKie]

As I mentioned: I have no knowledge of OpenCV. So don't know what they might mean with "scale".
Have you confirmed, that CV addresses from top-left to bottom-right?

[JorjMcKie]

Are you even sure that Figure 1 is an image at all?
Might as well be a drawing (PDF vector image instructions).

[uttam-scai]

I understand. Thank you.
In the example above I am not trying to create rectangle around Figure 1. I simply wanted to show that CV2 addresses the coordinate from TOP-LEFT. I am simply trying to create a rectangle from (1,1) to (100,100). and since both are starting from top left (blue rectangle) then both have to have same coordinate system for the page.. right?

I apologize for asking cv2 question here. Thank you for your help !!!