Fast encoding algorithms for vector quantization based on orthogonal transform

For vector quantization (VQ), it is extremely time-consuming to extract the similar codeword with input vector during the encoding process. In this paper, three efficient algorithms are proposed to extract the features of input vector using orthogonal transform, i.e., PCA transform, Hadamard transform, Haar wavelet transform, respectively. These features are then used to early remove impossible codeword in the distortion computations stage. From the experimental results, it is shown that the proposed approaches can largely decrease the computation time for achieving VQ coding with the same quality with full search algorithm. More specifically, compared with the DHSS algorithm, the proposed algorithm reduces the computational time by 31% to 61%. Compared with the Pan's algorithm, the proposed algorithm reduces the computational time by 62% to 75%. Compared with the Lai's algorithm, the proposed algorithm reduces the computational time by 48% to 58%. Compared with the HTPDE algorithm, the proposed algorithm reduces the computational time by 27% to 44%. Compared with the WTPDE algorithm, the proposed algorithm reduces the computational time by 21% to 45%. Moreover, the computation time of the HWT-based approach is less than all other previous algorithms.