Journal of Applied Sciences ›› 2006, Vol. 24 ›› Issue (2): 135-139.

• Articles • Previous Articles     Next Articles

A New Character-Localization Method in Wavelet Transform Domain

ZHANG Er-hu, ZHANG Xu-jin, ZHANG Zhi-gang   

  1. Department of Information Science, Xi'an University of Technology, Xi'an 710048, China
  • Received:2004-05-24 Revised:2004-07-21 Online:2006-03-31 Published:2006-03-31

Abstract: Characters in an image often provide important information of the image.Extraction of characterregions enables one to quickly browse and retrieve specific images.Based on analysis of the property of characters and study of their features in the wavelet transform domain, a new method for localizing and extracting characters is proposed.By using an adaptive dynamic threshold together with information of the adjacent blocks, the rate of correct-detection may be raised.A special feature of the method is that only partial decoding is required for encoded images involving wavelets such as JPEG2000 and MPEG4. Experimental results show that the processing is fast and correct-localization rate is high.

Key words: image retrieval, compressed-domain processing, wavelet transform, character localization

CLC Number: