Text in natural scene and video images contain a lot of information
which is useful for people’s daily life, such as the names of streets
and shops, traffic information, etc. In multi-lingual and multi-script
countries like India, information communication using multiple
languages/scripts is quite common. Text detection and recognition in
natural scene and video images is a challenging task due to variation
of low contrast, complex background, multiple scripts, etc.
Additionally, varying orientation of characters in a text line causes
the recognition performance to suffer more. We present an efficient
approach to detect and recognize multi-oriented text of Indic scripts
in scene and video images. Text detection will be performed using
spatial and texture analysis. For recognition, a path of sliding
window is estimated from text regions for feature extraction. Finally,
HMM-based technique will be applied for recognition. Preliminary
experiment shows that the proposed method is effective in Indic script
recognition.