ROB: Resolve some image extraction edge cases (#3371)

Relates to #3369.

Author: stefan6419846

pypdf/_page.py

@@ -601,12 +601,14 @@ def _get_ids_image(
         if ancest is None:
             ancest = []
         lst: List[Union[str, List[str]]] = []
-        if PG.RESOURCES not in obj or RES.XOBJECT not in cast(
-            DictionaryObject, obj[PG.RESOURCES]
+        if (
+                PG.RESOURCES not in obj or
+                is_null_or_none(resources := obj[PG.RESOURCES]) or
+                RES.XOBJECT not in cast(DictionaryObject, resources)
         ):
             return [] if self.inline_images is None else list(self.inline_images.keys())
 
-        x_object = obj[PG.RESOURCES][RES.XOBJECT].get_object()  # type: ignore
+        x_object = resources[RES.XOBJECT].get_object()  # type: ignore
         for o in x_object:
             if not isinstance(x_object[o], StreamObject):
                 continue

pypdf/_xobj_image_helpers.py

@@ -15,6 +15,7 @@
     EncodedStreamObject,
     NullObject,
     TextStringObject,
+    is_null_or_none,
 )
 
 if sys.version_info[:2] >= (3, 10):
@@ -54,7 +55,7 @@ def _get_imagemode(
         raise PdfReadError(
             "Color spaces nested too deeply. If required, consider increasing MAX_IMAGE_MODE_NESTING_DEPTH."
         )
-    if isinstance(color_space, NullObject):
+    if is_null_or_none(color_space):
         return "", False
     color_space_str: str = ""
     if isinstance(color_space, str):
@@ -156,6 +157,24 @@ def _extended_image_frombytes(
     return img
 
 
+def __handle_flate__indexed(color_space: ArrayObject) -> Tuple[Any, Any, Any, Any]:
+    count = len(color_space)
+    if count == 4:
+        color_space, base, hival, lookup = (value.get_object() for value in color_space)
+        return color_space, base, hival, lookup
+
+    # Deal with strange AutoDesk files where `base` and `hival` look like this:
+    #   /DeviceRGB\x00255
+    element1 = color_space[1]
+    element1 = element1 if isinstance(element1, str) else element1.get_object()
+    if count == 3 and "\x00" in element1:
+        color_space, lookup = color_space[0].get_object(), color_space[2].get_object()
+        base, hival = element1.split("\x00")
+        hival = int(hival)
+        return color_space, base, hival, lookup
+    raise PdfReadError(f"Expected color space with 4 values, got {count}: {color_space}")
+
+
 def _handle_flate(
     size: Tuple[int, int],
     data: bytes,
@@ -174,7 +193,7 @@ def _handle_flate(
     base: Any
     hival: Any
     if isinstance(color_space, ArrayObject) and color_space[0] == "/Indexed":
-        color_space, base, hival, lookup = (value.get_object() for value in color_space)
+        color_space, base, hival, lookup = __handle_flate__indexed(color_space)
     if mode == "2bits":
         mode = "P"
         data = bits2byte(data, size, 2)
@@ -365,7 +384,7 @@ def _get_mode_and_invert_color(
             if (
                 colors == 1
                 and (
-                    not isinstance(color_space, NullObject)
+                    not is_null_or_none(color_space)
                     and "Gray" not in color_space
                 )
             )

tests/test_images.py

@@ -572,3 +572,12 @@ def test_jbig2decode__jbig2globals():
 
     # Wrong image: 0.9618265964800714
     assert image_similarity(image.image, img) >= 0.999
+
+
+@pytest.mark.enable_socket
+def test_get_ids_image__resources_is_none():
+    url = "https://github.com/user-attachments/files/18381726/tika-957721.pdf"
+    name = "tika-957721.pdf"
+    reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))
+    page = reader.pages[2]
+    assert list(page.images.items()) == []

tests/test_xobject_image_helpers.py

@@ -1,5 +1,6 @@
 """Test the pypdf._xobj_image_helpers module."""
 from io import BytesIO
+from pathlib import Path
 
 import pytest
 
@@ -10,6 +11,10 @@
 
 from . import get_data_from_url
 
+TESTS_ROOT = Path(__file__).parent.resolve()
+PROJECT_ROOT = TESTS_ROOT.parent
+RESOURCE_ROOT = PROJECT_ROOT / "resources"
+
 
 @pytest.mark.enable_socket
 def test_get_imagemode_recursion_depth():
@@ -126,3 +131,32 @@ def test_extended_image_frombytes_zero_data():
 
     with pytest.raises(EmptyImageDataError, match="Data is 0 bytes, cannot process an image from empty data."):
         _extended_image_frombytes(mode, size, data)
+
+
+def test_handle_flate__autodesk_indexed():
+    reader = PdfReader(RESOURCE_ROOT / "AutoCad_Diagram.pdf")
+    page = reader.pages[0]
+    for name, image in page.images.items():
+        assert name.startswith("/")
+        image.image.load()
+
+    data = RESOURCE_ROOT.joinpath("AutoCad_Diagram.pdf").read_bytes()
+    data = data.replace(b"/DeviceRGB\x00255", b"/DeviceRGB")
+    reader = PdfReader(BytesIO(data))
+    page = reader.pages[0]
+    with pytest.raises(
+            PdfReadError,
+            match=r"^Expected color space with 4 values, got 3: \['/Indexed', '/DeviceRGB', '\\x00\\x80\\x00\\x80\\x80耀"  # noqa: E501
+    ):
+        for name, _image in page.images.items():  # noqa: PERF102
+            assert name.startswith("/")
+
+
+@pytest.mark.enable_socket
+def test_get_mode_and_invert_color():
+    url = "https://github.com/user-attachments/files/18381726/tika-957721.pdf"
+    name = "tika-957721.pdf"
+    reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))
+    page = reader.pages[12]
+    for _name, image in page.images.items():  # noqa: PERF102
+        image.image.load()