BUG: Inverse color in CMYK images (#1947)

Closes #1895

CMYK image without decode requires reverting scale (see PDF 1.7 spec, p243, §2, last sentence)

pypdf/filters.py

@@ -871,6 +871,25 @@ def _handle_jpx(
  elif lfilters is None:
  img, image_format, extension = Image.frombytes(mode, size, data), "PNG", ".png"
 
+ # CMYK image without decode requires reverting scale (cf p243,2§ last sentence)
+ decode = x_object_obj.get(
+ IA.DECODE, ([1.0, 0.0] * 4) if img.mode == "CMYK" else None
+ )
+ if (
+ isinstance(color_space, ArrayObject)
+ and color_space[0].get_object() == "/Indexed"
+ ):
+ decode = None # decode is meanless of Indexed
+ if decode is not None and not all(decode[i] == i % 2 for i in range(len(decode))):
+ lut: List[int] = []
+ for i in range(0, len(decode), 2):
+ dmin = decode[i]
+ dmax = decode[i + 1]
+ lut.extend(
+ round(255.0 * (j / 255.0 * (dmax - dmin) + dmin)) for j in range(256)
+ )
+ img = img.point(lut)
+
  if IA.S_MASK in x_object_obj: # add alpha channel
  alpha = _xobj_to_image(x_object_obj[IA.S_MASK])[2]
  if img.size != alpha.size:

tests/test_filters.py

@@ -366,8 +366,8 @@ def test_tiff_predictor():
 
 
 @pytest.mark.enable_socket()
-def test_cmyk():
- """Decode cmyk with transparency"""
+def test_rgba():
+ """Decode rgb with transparency"""
  url = "https://corpora.tika.apache.org/base/docs/govdocs1/972/972174.pdf"
  name = "tika-972174.pdf"
  reader = PdfReader(BytesIO(get_pdf_from_url(url, name=name)))
@@ -385,6 +385,29 @@ def test_cmyk():
  assert d < 0.01
 
 
+@pytest.mark.enable_socket()
+def test_cmyk():
+ """Decode cmyk"""
+ try:
+ from Crypto.Cipher import AES # noqa: F401
+ except ImportError:
+ return # the file is encrypted
+ url = "https://github.com/py-pdf/pypdf/files/11962229/DB-5368770_Vitocal_200-G.pdf"
+ name = "Vitocal.pdf"
+ reader = PdfReader(BytesIO(get_pdf_from_url(url, name=name)))
+ url_png = "https://user-images.githubusercontent.com/4083478/251283945-38c5b92c-cf94-473c-bb57-a51b74fc39be.jpg"
+ name_png = "VitocalImage.png"
+ refimg = Image.open(
+ BytesIO(get_pdf_from_url(url_png, name=name_png))
+ ) # not a pdf but it works
+ data = reader.pages[1].images[0]
+ diff = ImageChops.difference(data.image, refimg)
+ d = sqrt(
+ sum([(a * a + b * b + c * c + d * d) for a, b, c, d in diff.getdata()])
+ ) / (diff.size[0] * diff.size[1])
+ assert d < 0.01
+
+
 @pytest.mark.enable_socket()
 def test_iss1863():
  """Test doc from iss1863"""