todo

cluster.py

@@ -36,6 +36,7 @@ class Chars:
             result[-1].merge_in(char)
          else:
             result.append(Chars([char], self.page))
+      # TODO remove headers, footers; maybe median font size vs. max of cluster?
       result = [i for i in result if i.n > 2]
 
       # merge all vertically overlapping boxes