Wavelet based directional local extrema patterns for image retrieval on large image database

Abstract

Nowadays, images are part of our daily life, and content based image retrieval (CBIR) is a wide research area. Many researchers have done work on local patterns and directional patterns. In this paper, a new direction based feature descriptor has been proposed for content based image retrieval. The proposed method extracts the directional features using discrete wavelet transform (DWT) and directional local extrema patterns (DLEP). Wavelet coefficients are extracted from image, and based on direction information of the coeffi- cient, corresponding DLEP is applied. Presented method has been tested on Corel-5k and Corel-10k database, and evaluation measures are computed for the fruitfulness of this method. Comparison with earlier local pattern techniques indicates that this directional method is more significant than others.