get_page_images(): only return images actually referenced by the given page

src/__init__.py

@@ -29,6 +29,7 @@
 import warnings
 import weakref
 import zipfile
+from operator import itemgetter
 
 from . import extra
 
@@ -2923,6 +2924,8 @@ def __init__(self, filename=None, stream=None, filetype=None, rect=None, width=0
             self.is_encrypted = False
             self.is_encrypted = False
             self.metadata    = None
+            self.has_duplicate_images = False
+            self.images_xrefs_by_page = None
             self.FontInfos   = []
             self.Graftmaps   = {}
             self.ShownPages  = {}
@@ -3047,6 +3050,30 @@ def __init__(self, filename=None, stream=None, filetype=None, rect=None, width=0
                     self.page_count2 = extra.page_count_pdf
                 else:
                     self.page_count2 = extra.page_count_fz
+
+            # if the doc is a PDF, check for images duplication across pages
+            # this may happen, e.g., when converting from MS Office formats with external tools
+            # if the PDF has 1 page only, there is no possibility of duplication across pages
+            if self.is_pdf and self.page_count > 1:
+                has_duplicate_images = True
+                first_page_n_images = len(self.get_page_images(0))
+                for page in self.pages(start=1):
+                    # we need at least one page with a different number of images
+                    # to exclude full document duplication
+                    if len(page.get_images()) != first_page_n_images:
+                        has_duplicate_images = False
+                        break
+                self.has_duplicate_images = has_duplicate_images
+
+                if self.has_duplicate_images:
+                    self.images_xrefs_by_page = []
+                    for page in self.pages():
+                        # store only images referenced by page
+                        page_xrefs = list(map(
+                            itemgetter("xref"),
+                            page.get_image_info(xrefs=True)
+                        ))
+                        self.images_xrefs_by_page = page_xrefs
         finally:
             JM_mupdf_show_errors = JM_mupdf_show_errors_old
 
@@ -5090,7 +5117,14 @@ def get_page_images(self, pno: int, full: bool =False) -> list:
             return ()
         val = self._getPageInfo(pno, 2)
         if not full:
-            return [v[:-1] for v in val]
+            val = [v[:-1] for v in val]
+        if self.has_duplicate_images:
+            deduplicated_val = []
+            for v in val:
+                # v[0] is "xref"
+                if v[0] in self.images_xrefs_by_page[pno]:
+                    deduplicated_val.append(v)
+            return deduplicated_val
         return val
 
     def get_page_labels(self):