Abstract: In this paper，the texture similarity description of fabric structure image based on wavelet subband coefficients is presented.The Kullback-Leibler distance is computed to measure the texture similarity of fabric structure images using the scale parameter and shape parameter of the generalized Gaussian density，according to the fact that the subband histogram is fitted well using generalized Gaussian density.From the experiment results，note that the image of reference has a more significant effect on the curve of the Kullback-Leibler distance compared with the difference of wavelet base and decomposition level.Experimental results on 90 fabric texture image indicate that higher identification accuracy is gotten based on Kullback-Leibler distance compared with traditional approaches.The algorithms proposed in this paper can be applied to other kind’s texture.