cope with deflated images with CMYK Black Only

closes py-pdf#2321
pubpub-zz · Nov 29, 2023 · ec4e1ca · ec4e1ca
1 parent 441bb59
commit ec4e1ca
Show file tree

Hide file tree

Showing 3 changed files with 33 additions and 7 deletions.
diff --git a/pypdf/_xobj_image_helpers.py b/pypdf/_xobj_image_helpers.py
@@ -91,10 +91,18 @@ def _get_imagemode(
         )
         return mode2, True
     elif color_space[0] == "/DeviceN":
+        original_color_space = color_space
         color_components = len(color_space[1])
         color_space = color_space[2]
         if isinstance(color_space, IndirectObject):  # pragma: no cover
             color_space = color_space.get_object()
+        if color_space == "/DeviceCMYK" and color_components == 1:
+            if original_color_space[1][0] != "/Black":
+                logger_warning(
+                    f"Color {original_color_space[1][0]} converted to Gray. Please share PDF with pypdf dev team",
+                    __name__,
+                )
+            return "L", True
         mode2, invert_color = _get_imagemode(
             color_space, color_components, prev_mode, depth + 1
         )

diff --git a/pypdf/filters.py b/pypdf/filters.py
@@ -558,13 +558,16 @@ class CCITTFaxDecode:
 
     @staticmethod
     def _get_parameters(
-        parameters: Union[None, ArrayObject, DictionaryObject, IndirectObject], rows: int
+        parameters: Union[None, ArrayObject, DictionaryObject, IndirectObject],
+        rows: int,
     ) -> CCITParameters:
         # TABLE 3.9 Optional parameters for the CCITTFaxDecode filter
         k = 0
         columns = 1728
         if parameters:
-            parameters_unwrapped = cast(Union[ArrayObject, DictionaryObject], parameters.get_object())
+            parameters_unwrapped = cast(
+                Union[ArrayObject, DictionaryObject], parameters.get_object()
+            )
             if isinstance(parameters_unwrapped, ArrayObject):
                 for decode_parm in parameters_unwrapped:
                     if CCITT.COLUMNS in decode_parm:
@@ -778,8 +781,8 @@ def _xobj_to_image(x_object_obj: Dict[str, Any]) -> Tuple[Optional[str], bytes,
     alpha = None
     filters = x_object_obj.get(SA.FILTER, [None])
     lfilters = filters[-1] if isinstance(filters, list) else filters
-    if lfilters == FT.FLATE_DECODE:
-        img, image_format, extension, invert_color = _handle_flate(
+    if lfilters in (FT.FLATE_DECODE, FT.RUN_LENGTH_DECODE):
+        img, image_format, extension, _ = _handle_flate(
             size,
             data,
             mode,
@@ -821,15 +824,14 @@ def _xobj_to_image(x_object_obj: Dict[str, Any]) -> Tuple[Optional[str], bytes,
             ".png",
             False,
         )
-
     # CMYK image and other colorspaces without decode
     # requires reverting scale (cf p243,2§ last sentence)
     decode = x_object_obj.get(
         IA.DECODE,
         ([1.0, 0.0] * len(img.getbands()))
         if (
-            (img.mode == "CMYK" or (invert_color and img.mode == "L"))
-            and lfilters in (FT.DCT_DECODE, FT.JPX_DECODE)
+            (img.mode == "CMYK" and lfilters in (FT.DCT_DECODE, FT.JPX_DECODE))
+            or (invert_color and img.mode == "L")
         )
         else None,
     )

diff --git a/tests/test_images.py b/tests/test_images.py
@@ -220,3 +220,19 @@ def test_loop_in_image_keys():
     reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))
     reader.pages[0]["/Resources"]["/XObject"][NameObject("/toto")] = NullObject()
     reader.pages[0].images.keys()
+
+
+@pytest.mark.enable_socket()
+def test_devicen_cmyk_black_only():
+    """Cf #2321"""
+    url = "https://github.com/py-pdf/pypdf/files/13501846/Addressing_Adversarial_Attacks.pdf"
+    name = "iss2321.pdf"
+    reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))
+    url = "https://github.com/py-pdf/pypdf/assets/4083478/cc2dabc1-86e6-4179-a8a4-2b0efea124be"
+    name = "iss2321_img0.pdf"
+    img = Image.open(BytesIO(get_data_from_url(url, name=name)))
+    assert image_similarity(reader.pages[5].images[0].image, img) >= 0.99
+    url = "https://github.com/py-pdf/pypdf/assets/4083478/6b64a949-42be-40d5-9eea-95707f350d89"
+    name = "iss2321_img1.pdf"
+    img = Image.open(BytesIO(get_data_from_url(url, name=name)))
+    assert image_similarity(reader.pages[10].images[0].image, img) >= 0.99