An Extended LBG Algorithm for Constructing Universal Image Encoders

In recent years, principal component analysis (PCA) has attracted great attention in image compression. However, since the compressed image data include both the transformation matrix (the eigenvectors) and the transformed coefficients, PCA cannot produce the same performance as DCT (discrete Cosine transform) in respect of compression ratio. In using DCT, we need only to preserve the coefficients after transformation, because the transformation matrix is universal in the sense that it can be used to compress all images. To solve this problem we proposed k-PCA which is a potential universal image encoder. To increase the approximation ability of k-PCA, we propose an extended LBG (e-LBG) algorithm in this paper. The basic idea of the e-LBG algorithm is to improve the k-PCA iteratively using the training data. Experimental results show that the proposed approach is very effective, although the computing time is slightly increased. Key-Words: Image compression, PCA, vector quantization, k-PCA, universal encoder