Abstract: Conventional Gabor representation and its extracted features often yield a fairly poor performance in retrieving the rotated versions of the texture image under query. In this paper, a rotation-invariant Gabor texture representation is presented to address the issue. In the proposed algorithm, each representation only requires few summations on conventional Gabor filter impulse responses. The dimension of the new feature space is much reduced, which leads to a much smaller metadata storage space and faster on-line computation on the similarity measurement. Simulation results demonstrate that the proposed invariant Gabor representations and its extracted invariant features significantly are superior to conventional Gabor representation approach for retrieving the versions of rotated texture image.


Key words: content-based image retrieval, Gabor filter, rotation-invariant, texture analysis