Layout Analysis based on Text Line Segment Hypotheses

This paper describes on-going work in the development of a document layout analysis system based on text line segments. It describes two novel algorithms: gapped text line finding, which can identify text line segments, taking into account per-text line font information for the determination of where text line segments break, and reading order recovery for text line segments using topological sorting. An extension of the approach to a probabilistic maximum likelihood framework is discussed.