Multi-stage PCA image coding

Principal component analysis (PCA), a statistical processing technique, converts the data set to a lower dimensional feature space which reserve yet most of the internal information of the original data and reduces redundancy through the projection of data based on a proper orthonormal basis. PCA provides an efficient representation of the data compression. In this paper, we use Multi-stage PCA for image coding and partition the data set into some clusters with eigenvectors of a covariance matrix. Multi-stage clustering method can effectively remove redundancy and increase the numbers of principal components to improve the retrieved effect of certain clusters with complex structures. Using the proposed method, the retrieved image quality is relatively better and fewer storage variables are recorded.