METHOD OF FINDING COLUMNS IN TABULAR DOCUMENTS

Tabular documents have column structures that can be determined without decoding the bitmap. The method searches for separation intervals that separate word fragments in the table. These separation intervals are processed by intersecting them with other intervals and ranking the resulting intervals. A structured closure of separation intervals are maintained in bins. The intervals in the bins are sorted and used to determine new intersections when the next separation interval is processed. The intervals with the highest ranking are selected as the column separation intervals. The columns are easily identified with the method without first decoding the bitmap.