Croped and replaced Image Retuns Black

[daneshgit]

Please provide all mandatory information!

Describe the bug (mandatory)

I am applying redaction on the following pdf .In that pdf the image area is blacked out during the redaction process, as an alternate I am trying with the pixmap the get the original image without blacked out and replaced it after the redaction process on that particular image. Could be helpful is there is any alternate solution for removing the blacked
combined-pdf-92.pdf
out image.

To Reproduce (mandatory)

I given the following apply_redactions() , apply_redactions(images=1) , apply_redactions(images=2)
Following is the code I used:
basepix = page.getPixmap(clip=fitz.Rect(image['bbox']),alpha=False)
page.addRedactAnnot(image['bbox'], fill=colors.to_rgb("white"), text=" ")
page.apply_redactions()
page.insert_image(image['bbox'],pixmap=basepix,overlay=True,keep_proportion=True)

Expected behavior (optional)

I expect the same image original image.

Screenshots (optional)

If applicable, add screenshots to help explain your problem.
Output Image

Input Image

Two images overwritten example

Your configuration (mandatory)

• Operating system, potentially version and bitness
• Python version, bitness
• PyMuPDF version, installation method (wheel or generated from source).

For example, the output of print(sys.version, "\n", sys.platform, "\n", fitz.__doc__) would be sufficient (for the first two bullets).

Additional context (optional)

Add any other context about the problem here.

[JorjMcKie]

This a known upstream bug:
If you choose to whiten out areas of images which overlap redaction rectangles, then this will lead to undesired results if the image has a mask (i.e. an alpha channel). This is documented in the manual!

You have 2 options

1. do not touch overlapping images at all: fitz.PDF_REDACT_IMAGE_NONE
2. remove overlapping images completely: fitz.PDF_REDACT_IMAGE_REMOVE

So instead of your workaround just choose option 1 and do not let the method default to fitz.PDF_REDACT_IMAGE_PIXELS.
If you abslutely need to see whitened subareas of images that overlap, you must simulate the effect for alpha images by overlaying it with a redaction rectangle with a white background (use fill parameter).

[JorjMcKie]

Sorry, did not read your issue completely:
I tested your PDF by removing the word "Phone" using this snippet

rl = page.search_for("phone")
for r in rl:
    page.add_redact_annot(r)
page.apply_redactions(images=fitz.PDF_REDACT_IMAGES_NONE)

and so no black images at all.

[daneshgit]

Hi 'phone' is not an image hence its redacting.
I tried with "067024" which is located in the page end , the result is its not redacting using the below method.

rl = page.search_for("067024")
for r in rl:
page.add_redact_annot(r,fill=colors.to_rgb('red'))
page.apply_redactions(images=0)

As an alternative I got the coordinates of "067024" and redacted using the following method

1.This returns black image

        clip = fitz.Quad(((216,813), (257,813), (257,822), (216,822)))
        page.add_redact_annot(clip,fill=colors.to_rgb('red'))
        page.apply_redactions()

2.This returns like applying just a rectangle in the foreground but if we see in the background by editing pdf the keyword remains

        clip = fitz.Quad(((216,813), (257,813), (257,822), (216,822)))
        page.add_redact_annot(clip,fill=colors.to_rgb('red'))
        page.apply_redactions(images=0)

3.This returns like removing all images related to "067024"
clip = fitz.Quad(((216,813), (257,813), (257,822), (216,822)))
page.add_redact_annot(clip,fill=colors.to_rgb('red'))
page.apply_redactions(images=1)

4.This returns the Black image
clip = fitz.Quad(((216,813), (257,813), (
257,822), (216,822)))
page.add_redact_annot(clip,fill=colors.to_rgb('red'))
page.apply_redactions(images=2)

Can I have the solution to remove "067024" without affecting anything others

[JorjMcKie]

Can I have the solution to remove "067024" without affecting anything others

No, not "<0677024>", because this is part of an image. You won't find string "<0677024>".
String "0677024" does get removed fine (inside the box "Office use only"). That text has been placed above the image and can therefore be redacted.

[daneshgit]

Hi , As Its is in the document apply_redaction(images=2) removes only the overlapping part of the image on which redaction is annotated. May I know why its blacked out all.

Input image with the following code:
clip = fitz.Quad(((216,813), (257,813), (257,822), (216,822)))
page.add_redact_annot(clip,fill=colors.to_rgb('white'))

test18.pdf

Result after applying the following code
apply_redaction(images=2)
test20.pdf

[JorjMcKie]

It also is documented, that this part of MuPDF code is not bug free. For images with a mask (transparency present), and also some black and white images, that logic is not (always) working.
They know about the bug.

[agsdineshDev]

Whoever fixes the transparent image apply reaction issues and the layered image issue described above will be reward with bitcoin worth 1000 Australian Dollar.
Share the fixed python code and bitcoin address with us once upon verification. The reward will be processed.

Bag your reward.

Good luck!!!

[JorjMcKie]

Interesting offer!
I have developed something which results in this for your example:
curate-images.pdf
I do not want to attach my code in this thread: if you are serious, then I am serious as well.
Please let me know how to proceed further 😎