Fix: MinerU parsed images are obscured when using MinerU as a parser.…

deepdoc/parser/mineru_parser.py

@@ -358,7 +358,6 @@ def _line_tag(self, bx):
         return "@@{}\t{:.1f}\t{:.1f}\t{:.1f}\t{:.1f}##".format("-".join([str(p) for p in pn]), x0, x1, top, bott)
 
     def crop(self, text, ZM=1, need_position=False):
-        imgs = []
         poss = self.extract_positions(text)
         if not poss:
             if need_position:
@@ -416,7 +415,12 @@ def crop(self, text, ZM=1, need_position=False):
         )
 
         positions = []
+        imgs = []
+        head_ctx_end = 0
+        tail_ctx_start = 0
         for ii, (pns, left, right, top, bottom) in enumerate(poss):
+            if ii + 1 == len(poss):
+                tail_ctx_start = len(imgs)
             right = left + max_width
 
             if bottom <= top:
@@ -468,6 +472,9 @@ def crop(self, text, ZM=1, need_position=False):
                     positions.append((pn + self.page_from, x0, x1, y0, y1))
                 bottom -= page.size[1]
 
+            if ii == 0:
+                head_ctx_end = len(imgs)
+
         if not imgs:
             if need_position:
                 return None, None
@@ -481,7 +488,7 @@ def crop(self, text, ZM=1, need_position=False):
         pic = Image.new("RGB", (width, height), (245, 245, 245))
         height = 0
         for ii, img in enumerate(imgs):
-            if ii == 0 or ii + 1 == len(imgs):
+            if ii < head_ctx_end or ii >= tail_ctx_start:
                 img = img.convert("RGBA")
                 overlay = Image.new("RGBA", img.size, (0, 0, 0, 0))
                 overlay.putalpha(128)

test/unit_test/deepdoc/parser/test_mineru_crop.py

@@ -0,0 +1,128 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+"""Unit tests for MinerUParser.crop() overlay logic.
+
+Regression tests for issue #14197: content images were incorrectly darkened
+when the top context strip had zero height and was skipped.
+"""
+
+import logging
+
+import pytest
+from PIL import Image
+
+from deepdoc.parser.mineru_parser import MinerUParser
+
+
+def _make_parser(page_images):
+    parser = MinerUParser.__new__(MinerUParser)
+    parser.page_images = page_images
+    parser.page_from = 0
+    parser.logger = logging.getLogger("test_mineru_crop")
+    return parser
+
+
+def _solid_image(color, size=(500, 800)):
+    return Image.new("RGB", size, color)
+
+
+def _sample_center(img):
+    w, h = img.size
+    return img.getpixel((w // 2, h // 2))
+
+
+class TestCropOverlay:
+    @pytest.mark.p2
+    def test_content_not_darkened_when_top_context_skipped(self):
+        """
+        Bug #14197: image at top of page (top=0) produces a zero-height top
+        context strip that is skipped. The content image must NOT be darkened.
+        """
+        RED = (255, 0, 0)
+        parser = _make_parser([_solid_image(RED)])
+
+        # top=0 → context strip above has zero height → skipped
+        tag = "@@1\t50.0\t450.0\t0.0\t200.0##"
+        result = parser.crop(tag)
+
+        assert result is not None
+        r, g, b = _sample_center(result)
+        # darkened overlay would halve the red channel to ~127
+        assert r > 200, f"Content image darkened (r={r}); overlay incorrectly applied to content"
+
+    @pytest.mark.p2
+    def test_content_not_darkened_when_image_near_top(self):
+        """Image within GAP(6px) of page top also produces zero-height top strip."""
+        RED = (255, 0, 0)
+        parser = _make_parser([_solid_image(RED)])
+
+        # top=4 → max(4-6, 0)=0 → zero-height top context strip
+        tag = "@@1\t50.0\t450.0\t4.0\t200.0##"
+        result = parser.crop(tag)
+
+        assert result is not None
+        r, g, b = _sample_center(result)
+        assert r > 200, f"Content image darkened (r={r}); overlay incorrectly applied to content"
+
+    @pytest.mark.p2
+    def test_context_strips_are_darkened(self):
+        """Context strips above and below content must receive the overlay."""
+        WHITE = (255, 255, 255)
+        parser = _make_parser([_solid_image(WHITE, size=(500, 800))])
+
+        # Content in the middle of the page — both context strips are valid
+        tag = "@@1\t50.0\t450.0\t300.0\t400.0##"
+        result = parser.crop(tag)
+
+        assert result is not None
+        # Top-most pixel row should be from the darkened top context strip
+        r, g, b = result.getpixel((result.size[0] // 2, 0))
+        assert r < 200, f"Top context strip not darkened (r={r})"
+
+    @pytest.mark.p1
+    def test_single_image_not_darkened_when_both_context_strips_skipped(self):
+        """
+        Core bug from #14197: when both context strips are skipped (len(imgs)==1),
+        the single content image must NOT receive the overlay.
+        Original code: ii==0 AND ii+1==len(imgs) both True → always darkened.
+        """
+        RED = (255, 0, 0)
+        # Page height = 200px; content fills full page → both context strips zero-height
+        parser = _make_parser([_solid_image(RED, size=(500, 200))])
+
+        # top=0, bottom=200 = page height → both context strips are zero-height → skipped
+        tag = "@@1\t50.0\t450.0\t0.0\t200.0##"
+        result = parser.crop(tag)
+
+        assert result is not None
+        r, g, b = _sample_center(result)
+        # Before fix: r≈127 (darkened). After fix: r≈255 (clear).
+        assert r > 200, f"Single content image darkened (r={r}); both-strips-skipped bug not fixed"
+
+    @pytest.mark.p2
+    def test_multi_page_content_not_darkened(self):
+        """Content spanning multiple pages must not be darkened."""
+        RED = (255, 0, 0)
+        parser = _make_parser([_solid_image(RED), _solid_image(RED)])
+
+        # top=0 on page 1, spans to page 2
+        tag = "@@1-2\t50.0\t450.0\t0.0\t100.0##"
+        result = parser.crop(tag)
+
+        assert result is not None
+        r, g, b = _sample_center(result)
+        assert r > 200, f"Multi-page content image darkened (r={r})"