Paper-cutting Image Retrieval Technology based on LSH Improvement

Abstract

Characteristics of paper-cutting, paper-cutting gallery construction, etc. are analyzed aiming at existing problems of folk traditional paper-cutting art multimedia interactive platform. It is proposed that rotation invariant LBP (Local Binary Pattern) is combined with LSH algorithm to present a large-scale paper-cutting image fast retrieval method. Firstly, the background image is eliminated with OTSU, then the paper-cutting image rotation LBP feature. In addition, highdimensional data is mapped to low dimensional space, and a hash index was constructed by local sensitive hash algorithm to find the approximate KNN. Experimental result in the dataset shows that the improved algorithm has high accuracy on paper-cutting image retrieval, and it is significantly higher than traditional algorithm on retrieval speed.