Cropping investigation

copienator/edge_marks.py

@@ -0,0 +1,95 @@
+"""Group small isolated border marks using evidence from one displayed page.
+
+This identifies dispensable border marks, without recognizing pagination or
+consulting other pages. Component IDs are local to the caller's foreground.
+Long paper-line candidates are excluded from grouping but remain in the crop.
+"""
+import numpy as np
+
+
+def classify_edge_marks(labels, stats, eligible, rgb, dpi):
+    px = dpi/25.4
+    height, width = labels.shape
+    components = []
+    for i, (x, y, w, h, area) in enumerate(stats[1:], 1):
+        if not eligible[i] or area < .15*px*px:
+            continue
+        # Exclusion from candidate grouping ONLY. Pixels remain in the crop mask.
+        if ((h > 25*px and w < 3*px) or (w > 25*px and h < .6*px)):
+            continue
+        components.append(i)
+    parent = {i:i for i in components}
+    def find(i):
+        while parent[i] != i:
+            parent[i] = parent[parent[i]]
+            i = parent[i]
+        return i
+    # A y sweep bounds the search on noisy pages; uncertain large groups stay.
+    edge_components = [i for i in components if stats[i,2] <= 40*px
+                       and stats[i,3] <= 18*px and
+                       (stats[i,1]+stats[i,3] <= 40*px
+                        or stats[i,1] >= height-40*px)]
+    if len(edge_components) > 512:
+        return [], []
+    colours = {}
+    if rgb is not None:
+        for i in edge_components:
+            x,y,w,h,_ = stats[i]
+            values = rgb[y:y+h,x:x+w][labels[y:y+h,x:x+w]==i]
+            colours[i] = float(np.median(values.max(axis=1)-values.min(axis=1)))
+    ordered = sorted(edge_components, key=lambda i:stats[i,1])
+    for a, i in enumerate(ordered):
+        x,y,w,h,_ = stats[i]
+        for j in ordered[a+1:]:
+            xx,yy,ww,hh,_ = stats[j]
+            if yy > y+h+1.5*px:
+                break
+            gap_x = max(0, xx-(x+w), x-(xx+ww))
+            gap_y = max(0, yy-(y+h), y-(yy+hh))
+            incompatible = (rgb is not None and
+                ((colours[i] < 35 and colours[j] > 60)
+                 or (colours[j] < 35 and colours[i] > 60)))
+            if not incompatible and gap_x <= 3*px and gap_y <= 1.5*px:
+                parent[find(j)] = find(i)
+    groups = {}
+    for i in edge_components:
+        groups.setdefault(find(i), []).append(i)
+    rows = []
+    for ids in groups.values():
+        s=stats[ids];x,y=s[:,:2].min(axis=0);right,bottom=(s[:,:2]+s[:,2:4]).max(axis=0)
+        w,h=right-x,bottom-y
+        if not (w <= 40*px and h <= 18*px and h >= 1.5*px):
+            continue
+        edge = 'top' if bottom <= 40*px else 'bottom' if y >= height-40*px else None
+        if edge is None:
+            continue
+        underline_ids=[]
+        for i in ids:
+            xx,yy,ww,hh,area=stats[i]
+            if ww > 3*hh and ww >= .65*w and yy+hh >= bottom-1.5*px:
+                underline_ids.append(i)
+        lateral = (x+w/2 < width*.25 or x+w/2 > width*.75)
+        # Check separation from other compact strokes across the full width.
+        # Confirmed long line candidates remain in the actual crop mask.
+        body = [i for i in components if i not in ids and stats[i,3] < 18*px]
+        edge_distance = (y if edge == 'top' else height-bottom)/px
+        if body:
+            gap = ((min(stats[i,1] for i in body)-(y+h))/px if edge=='top' else
+                   (y-max(stats[i,1]+stats[i,3] for i in body))/px)
+        else:
+            gap = None
+        truncated = (x <= px or y <= px or right >= width-px or bottom >= height-px)
+        reasons=[]
+        if lateral:reasons.append('lateral-placement')
+        if gap is not None and gap >= 20:reasons.append('all-column-stroke-gap>=20mm')
+        if edge_distance <= 15:reasons.append('within-15mm-of-edge')
+        if underline_ids:reasons.append('underline-protection')
+        if truncated:reasons.append('scan-edge-truncated')
+        proposed = (lateral and edge_distance <= 15 and gap is not None
+                    and gap >= 20 and not underline_ids and not truncated)
+        rows.append(dict(bbox_mm=[round(float(v/px),2) for v in [x,y,x+w,y+h]],
+                         component_ids=ids,edge=edge,edge_distance_mm=round(edge_distance,2),gap_mm=None if gap is None else round(gap,2),
+                         evidence=reasons,proposed_pagination=bool(proposed)))
+    removed = [i for row in rows if row['proposed_pagination'] for i in row['component_ids']]
+    return removed, rows
+