Text Orientation Detection from Document Image of Indian Scripts

This paper describes an effective and robust technique to determine orientation of text perceiving in document image as well as restoring it to right orientation. Described technique is useful in scripts of Hindi, Bengali, and Punjabi languages. It is based on pattern recognition in document. Specific pattern has been decided by extensive study of different scripts. On the basis of specific pattern, area of text of interest in document is determined and after analysis of these text area, orientation and skew of the document is determined. After that, document is restored to right orientation by rotating the document in reverse direction. This technique is also useful in those documents where text is coming with images, and document is in multicolumn. Orientation is determined by partitioning whole document into sub blocks of fixed size. Histogram Analysis of text pattern is done on each sub blocks to find orientation of document.