pdf to image rendering ignore optional content offs

[PasaOpasen]

Description of the bug

with some docs with already disabled optional content layers the rendered pages still contain them;

example link: https://dropmefiles.com/zTbp4

How to reproduce the bug

f = 'path/oc-for-ocr.pdf'
dpi = 150

import fitz
from PIL import Image

doc = fitz.open(f)
print(doc.layer_ui_configs())  # shows that almost all layers except "Text" are off

pix = doc[0].get_pixmap(matrix=fitz.Matrix(dpi/72, dpi/72))
img =Image.frombytes('RGB', (pix.width, pix.height), pix.samples)

img.show()  # displays image containing all layers info

PyMuPDF version

1.24.9

Operating system

Windows

Python version

3.8

[JorjMcKie]

Cannot download from the supplied link. Please provide a working one.

[PasaOpasen]

@JorjMcKie sent to your email

[JorjMcKie]

This is an upstream error. Opening a MuPDF report. Is this a confidential file, or can I simply attach it here?

[PasaOpasen]

@JorjMcKie sorry, it is confidential

is there any way to hotfix the problem? Like removing hidden layers content

[PasaOpasen]

pypdfium2 encounters same problem on this doc but pdf2image and python-poppler work well

[JorjMcKie]

A similar result from all the browsers: some show Text only, some show also other content.

[JorjMcKie]

File link: https://drive.google.com/file/d/10dREl0VwJx9HK8W8m3RLmu2sECMrgxbC/view?usp=drive_link

Bugzilla link: https://bugs.ghostscript.com/show_bug.cgi?id=708268

[julian-smith-artifex-com]

[As of 2025-3-12, the upstream bug https://bugs.ghostscript.com/show_bug.cgi?id=708268 has not been fixed.]

[alexandrebouillet]

I encountered the same issue where doc.set_layer() doesn't properly disable layers when PDF files contain OCG (Optional Content Groups) with explicit Usage properties like:

{'/Name': 'Layer Name', '/Type': '/OCG', '/Usage': {'/View': {'/ViewState': '/ON'}, '/Print': {'/PrintState': '/ON'}}}

Problem

When OCGs have predefined ViewState and PrintState in their Usage dictionary, set_layer(-1, on=on, off=off) only modifies the layer visibility state but doesn't update the Usage properties. This causes layers to remain visible during PDF-to-image conversion even when marked as "off".

Workaround Solution

I found that manually updating the Usage properties via xref_set_key() before calling set_layer() resolves the issue:

  def disable_pdf_layers(pdf_path: str, disabled_layers: list[str], enabled_layers: list[str]) -> None:
      doc = fitz.open(pdf_path)
      ocgs = doc.get_ocgs().items()

      on, off = [], []
      enabled_mode = bool(enabled_layers)

      for xref, ocg in ocgs:
          name = ocg.get("name", "Unnamed").lower()

          # Determine if layer should be on/off
          if enabled_mode:
              should_be_on = any(layer in name for layer in enabled_layers)
              (on if should_be_on else off).append(xref)
          else:
              should_be_on = not any(layer in name for layer in disabled_layers)
              (on if should_be_on else off).append(xref)

          # Update Usage properties directly via xref
          xref_object = doc.xref_object(xref)
          state_name = "ON" if should_be_on else "OFF"

          if "ViewState" in xref_object:
              doc.xref_set_key(xref, "Usage/View/ViewState", f"/{state_name}")
          if "PrintState" in xref_object:
              doc.xref_set_key(xref, "Usage/Print/PrintState", f"/{state_name}")
          if "ExportState" in xref_object:
              doc.xref_set_key(xref, "Usage/Export/ExportState", f"/{state_name}")

      # Keep original set_layer functionality
      doc.set_layer(-1, on=on, off=off)
      doc.save(pdf_path, incremental=True, encryption=0)
      doc.close()

This workaround successfully resolves the layer visibility issues during PDF processing and image conversion.

[sebras]

Bugzilla link: https://bugs.ghostscript.com/show_bug.cgi?id=708268

The upstream bug has been fixed.

[JorjMcKie]

The fix will be available in one of the next PyMuPDF versions.