Efficient histogram for region based image retrieval in the discrete cosine transform domain

Amina Belalia, Kamel Belloulata, Shiping Zhu

Abstract


Recently, several approaches of content based-image retrieval (CBIR), based on the characteristics of discrete cosine transform (DCT), such as decorrelation and concentration of energy in only a few coefficients, have been proposed. To reduce the semantic gap between high level search and low level patterns, a new concept based on region based search region-based image retrieval (RBIR) has been proposed. Recently, we proposed to use shape-adaptive (SA) DCT in a new RBIR system. In this paper, we propose an efficient histogram optimization suited to our region-based concept. This histogram takes into account the pattern’s from the SA-DCT of the border blocks as well as the DCT coefficients of the internal blocks. Our proposed scheme has greatly improved the results compared to region-based reference methods. Regionbased search is limited to the object of interest only, i.e. a region of the query image can only match a region of another image in the database.


Keywords


CBIR; Discrete cosine transform; RBIR; Semantic retrieval; Shape-Adaptive-DCT

Full Text:

PDF


DOI: http://doi.org/10.11591/ijai.v11.i2.pp546-563

Refbacks

  • There are currently no refbacks.


View IJAI Stats

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.