Texture feature extraction method for scale and rotation invariant image retrieval

Abstract

In content-based image retrieval systems, the texture in the query image may appear at a different scale and rotation angle to the relevant images in the database. To overcome this limitation, proposed is a new method to extract features that are invariant to the scale and rotation of the texture in the image. The proposed method uses circular shifting in both the scale and the rotation dimensions in conjunction with a Gaussian window to reduce the effect of very high and very low frequencies in the feature matching process. © 2012 The Institution of Engineering and Technology.