Information technologies Subband SD And Kurtosis Featured NSST Texture Retrieval

The selection of features extracted for image retrieval is quite important. We in this paper choose standard deviation and kurtosis to character texture. Non-subsampled shearlet transform is used to derive texture features. Shearlet is a new sparse representation tool of multidimensional function, which provides a simple and efficient mathematical framework. We firstly decompose the source images on various scales and in different directions with non-subsampled shearlet transform. The standard deviation and kurtosis of each subband are cascaded as texture features of the images. The similarity measurement of the image retrieval system is achieved with the Average Euclidean Distance. The experiments are executed on four different multidimensional transforms: DWT, Contourlet, NSCT and NSST. The results show the proposed retrieval method is superior to that of the traditional methods.