Automatic logo based document image retrieval

Subtitle:

Article type: Research Article

Authors: Shirdhonkar, M.S.^{a; *} | Kokare, Manesh B.^b

Affiliations: [a] BLDEA's Dr. P.G.H. CET, Bijapur, Karnataka, India | [b] SGGS Institute of Technology and Engineering, Nanded, Maharashtra, India

Correspondence: [*] Corresponding author: M.S. Shirdhonkar, BLDEA's Dr. P.G.H. CET, Bijapur, Karnataka, India. E-mail:ms_shirdhonkar@rediffmail.com

Abstract: Recently, there is need for the automatic logo detection and document image retrieval because of increasing requirements of intelligent document images. In this work, we have developed an automatic logo based document image retrieval system in which we first proposed an automatic logo detection and extraction from the document images using discrete wavelet transform (DWT). Then, we have proposed an approach to retrieve document images based on the logo using combined rotated complex wavelet filters (RCWF) and dual tree complex wavelet transform (DT-CWT). Since these combined features give information in twelve different directions. Our approach is segmentation free. Finally, we evaluate the effectiveness of our approach using large database collections of real-world complex documents.

Keywords: Document image retrieval, document similarity measurements, logo detection, wavelet transforms

DOI: 10.3233/KES-150305

Journal: International Journal of Knowledge-based and Intelligent Engineering Systems, vol. 19, no. 1, pp. 27-33, 2015