Abstract:

In order to manage and query massive images, there is an urgent need for a content

based image retrieval system with high retrieval rates. This paper proposes an image retrieval

method based on curvelet transform and integrated Shannon entropy and frequency subband

energy features. This method uses the Shannon entropy to preclassify data, adopts the energy

subband features to measure image similarity, and joins the feedback information of

searchers  to achieve accurate image retrieval. The retrieval experimental results conducted

on the Brodatz texture image database show that the system has a high retrieval rate and a

certain practical value.

Key words: texture retrieval;curvelet filtering;Shannon entropy