ROB: Resolve UnboundLocalError for xobjs in _get_image (#3684)

The xobjs variable was used outside the try-except block that defined it.
If a KeyError was caught and the id started/ended with '~' (inline images),
the code would continue but xobjs would remain undefined, causing
UnboundLocalError when trying to access non-inline images.

Initialize xobjs to None and check before using it.

---------

Co-authored-by: Yuki9814 <Yuki9814@users.noreply.github.com>
Co-authored-by: Yuki9814 <222397878+Yuki9814@users.noreply.github.com>

Author: Yuki9814

pypdf/_page.py

@@ -637,21 +637,25 @@ def _get_image(
             id = list(id)
         if isinstance(id, list) and len(id) == 1:
             id = id[0]
+        xobjs: Optional[DictionaryObject] = None
         try:
             xobjs = cast(
                 DictionaryObject, cast(DictionaryObject, obj[PG.RESOURCES])[RES.XOBJECT]
             )
-        except KeyError:
+        except KeyError as exc:
             if not (id[0] == "~" and id[-1] == "~"):
-                raise
+                raise KeyError(
+                    f"Cannot access image object {id} without XObject resources"
+                ) from exc
         if isinstance(id, str):
             if id[0] == "~" and id[-1] == "~":
                 if self.inline_images is None:
                     self.inline_images = self._get_inline_images()
-                if self.inline_images is None:  # pragma: no cover
+                if self.inline_images is None:
                     raise KeyError("No inline image can be found")
                 return self.inline_images[id]
 
+            assert xobjs is not None
             from .generic._image_xobject import _xobj_to_image  # noqa: PLC0415
             imgd = _xobj_to_image(cast(DictionaryObject, xobjs[id]))
             extension, byte_stream = imgd[:2]
@@ -662,6 +666,7 @@ def _get_image(
                 indirect_reference=xobjs[id].indirect_reference,
             )
         # in a subobject
+        assert xobjs is not None
         ids = id[1:]
         return self._get_image(ids, cast(DictionaryObject, xobjs[id[0]]))
 

tests/test_images.py

@@ -227,6 +227,34 @@ def test_image_extraction(src, page_index, image_key, expected):
     assert image_similarity(BytesIO(actual_image.data), expected) >= 0.99
 
 
+def test_get_inline_image_without_xobject_resources():
+    page = PageObject(None, None)
+    inline_image = object()
+
+    with mock.patch.object(page, "_get_inline_images", return_value={"~0~": inline_image}):
+        assert page._get_image("~0~") is inline_image
+
+
+def test_get_inline_image_without_xobject_resources_raises_when_missing():
+    page = PageObject(None, None)
+
+    with (
+        mock.patch.object(page, "_get_inline_images", return_value=None),
+        pytest.raises(KeyError, match="No inline image can be found"),
+    ):
+        page._get_image("~0~")
+
+
+def test_get_xobject_image_without_xobject_resources_raises():
+    page = PageObject(None, None)
+
+    with pytest.raises(
+        KeyError,
+        match="Cannot access image object /Im0 without XObject resources",
+    ):
+        page._get_image("/Im0")
+
+
 @pytest.mark.enable_socket
 @pytest.mark.timeout(30)
 def test_loop_in_image_keys():