Ancient-coins images retrieve by wavelet transform and relative moments

Abstract

A novel method of ancient coins feature extraction and image retrieve by wavelet transform and the relative moments is proposed, which firstly utilized the multi-scale wavelet feature template to achieve the polish output of color filter component and to realize gradient vector extended. Through multi-scale Wavelet transform, the boundary images at different scales of the color images of ancient coins were reaped; secondly, through the extracting of regional characteristics of each scale image based on the ten relative moments which were transformed from seven invariant moments, feature vectors of the image were formed; finally, feature vectors were normalized and its similarity was measured and then an image retrieval system was built. Experimental results show that the method has good versatility and robustness. © 2013 Springer-Verlag.