Rotation invariant texture spectrum-based image retrieval algorithm

Abstract

Image retrieval is to find similar images in image library, LBP texture spectrum is sensitive to the rotation for this problem, a primitive rotating LBP texture spectrum obtained after a certain point of view as a new texture spectrum consistent with the visual texture primitive spectrum, will have all the texture spectrum of visual consistency primitives into one, this can reduce the dimension of feature values, and image rotation can solve the problem of similarity calculation, using the texture spectrum for texture image description, and finally histogram with the texture primitive search, the results show that the precision than the traditional LBP has significantly improved, better search results.