Summary and Conclusion

The task of script segmentation in a multi-script document is a challenging problem. We have used a Gabor filter based multi-channel filtering technique to solve the problem. The various scripts in the document are considered to define different textures and a texture segmentation based approach has been used for the segmentation. The algorithm has been tested on various textures and multi-script documents and the results obtained are encouraging. The documents have been obtained from various Indian web sites containing newspaper articles in Indian scripts. The time requirement depends on the the number of filters used. Although, the approach has been tried only on textures and Indian scripts, it is very general and can easily be applied to other scripts. The approach will also work for documents where text and graphics are intermingled (example advertisements).