Implement proper diacritic detection.

pixelocr/page.py

@@ -15,6 +15,7 @@
 
 
 import itertools
+from collections import defaultdict
 
 import numpy as np
 from scipy import ndimage
@@ -129,21 +130,27 @@ class Line(PageObject):
         return margins.min()
 
     def _combine_diacritics(self, glyphs):
-        def is_diacritic(glyph):
-            # XXX
-            return (
-                glyph is not None
-                and glyph is not None
-                and glyph.elevation > 5
-            )
 
-        while glyphs:
-            glyph, *glyphs = glyphs
-            diacritics = list(itertools.takewhile(is_diacritic, iter(glyphs)))
-            for diacritic in diacritics:
-                glyph = Glyph(glyph.image.combine(diacritic.image), elevation=glyph.elevation)
-            glyphs = glyphs[len(diacritics):]
-            yield glyph
+        def find_diacritics(glyphs):
+            bodies = defaultdict(list)
+            diacritics = defaultdict(list)
+            for i, glyph in enumerate(glyphs):
+                if not glyph.is_body():
+                    continue
+                neighbours = glyphs[i - 5: i] + glyphs[i + 1: i + 6]
+                for neighbour in neighbours:
+                    if glyph.detect_diacritic(neighbour):
+                        diacritics[glyph].append(neighbour)
+                        bodies[neighbour].append(glyph)
+            return bodies, diacritics
+
+        bodies, diacritics = find_diacritics(glyphs)
+        for glyph in glyphs:
+            if glyph.is_body():
+                yield glyph.add_diacritics(*diacritics[glyph])
+            else:
+                if glyph not in bodies: # freestanding diacritic-like glyphacter without a body
+                    yield glyph
 
     def _insert_spaces(self, glyphs):
         for glyph, next_glyph in pairwise(glyphs):
@@ -161,6 +168,10 @@ class Line(PageObject):
 
 
 class Glyph(PageObject):
+    DIACRITIC_WINDOW_LEFT = 3
+    DIACRITIC_WINDOW_RIGHT = 5
+    DIACRITIC_MIN_ELEVATION = 5
+
     def __init__(self, image, elevation):
         super().__init__(image)
         self.elevation = elevation
@@ -170,6 +181,24 @@ class Glyph(PageObject):
         """Return a dictionary key uniquely representing this glyph."""
         return self.elevation, self.image.tostring()
 
+    def is_body(self):
+        """Return True if the glyph is definitely not diacritic."""
+        return self.elevation <= 0
+
+    def detect_diacritic(self, glyph):
+        """Return True if the given glyph can be our diacritic."""
+        return (
+            glyph.elevation >= self.DIACRITIC_MIN_ELEVATION
+            and glyph.left >= self.left - self.DIACRITIC_WINDOW_LEFT
+            and glyph.right <= self.right + self.DIACRITIC_WINDOW_RIGHT
+        )
+
+    def add_diacritics(self, *diacritics):
+        if not diacritics:
+            return self
+        if len(diacritics) > 1:
+            raise NotImplementedError
+        return Glyph(self.image.combine(diacritics[0].image), self.elevation)
 
 
 class Space(Glyph):