Abstract: A novel image Hashing based on principal component analysis (PCA) was proposed. PCA was introduced to reduce dimension of samples, and the projection matrix was achieved by choosing several eigenvectors which have higher recognition ability. Based on which, the reduced-sample was mapped with locality preserving projection (LPP). Meanwhile, the projection matrix of principal component analysis was randomly rotated to form a series of transformational matrixes. The matrix stitching was adopted to construct the final code projection matrix. Finally, the original samples were projected into the code projection matrix to get a reduced dimensional sample, and the Hashing code was used to achieve the final binary encoding. Experiments show that the proposed method has better stability, lower memory consumption and higher efficiency compared with other traditional methods.

Key words: Hashing, principal component analysis, locality preserving projection, image processing