Texture Retrieval Based on Nonlocal Singular Value Decomposition and Multiscale Transforms
J.W. Zhang, Bingqi Liu · Advances in engineering research/Advances in Engineering Research · 2015
Feature extraction based on the multiscale transforms is usually accomplished in the high frequency subbands.The features which are captured from the high frequency subbands mainly contain the edge changing information of images; meanwhile the structural information of images is mainly existed in the low frequency subband.In order to improve the effect of image processing, the structural features which are very important should be obtained from images.This paper adopts a nonlocal singular value decomposition (NL-SVD) algorithm to extract structural features from the low frequency subband.The structural features are combined with traditional statistical features such as means and standard deviations.These features are applied for texture retrieval.Experimental results show the retrieval ratio with these combined features is better and it proves that structural features are efficient texture features for images.