Vector quantization of still images using reflected subcodevectors

Vector quantization (VQ) is popularly used for image compression and many variants of VQ have been found to be successful. A codebook comprising codevectors of a given dimension exploits the linear and non-linear dependencies in an image. This paper proposes a novel method of designing a codebook using reflected subcodevectors. This scheme assumes that an image vector with reflected subvectors can be similar to another image vector. The proposed scheme considers such orientations of subcodevectors for exploiting the redundancy in an image. Results are presented for traditional VQ and the proposed scheme using standard test images.