Texture Image Retrieval System Based on Non-Subsampled Contourlet and Generalized Gaussian Density Model

Abstract

Aiming at improving the retrieval rate of the original contour let transform based texture image retrieval system, a non-sub sampled contour let transform based texture image retrieval system was proposed. Generalized Gaussian Density (GGD) model parameters were cascaded to form feature vectors and Kullback-Leibler distance (KLD) function was used for similarity measure. Experimental results on 640 texture images from Vistex texture image database indicate that non-sub sampled contour let transform based image retrieval system is superior to that of the original contour let transform under the same system structure with almost same length of feature vectors, retrieval time and memory needed. Furthermore, GGD combined with KLD method has higher retrieval rates than energy based features combined with Euclidean distance under comparable levels of computational complexity, decomposition parameters including the number of scale and directional sub band on each scale selected in both contour let transforms can make effects on retrieval rates.