An Adaptive Two-Stage Codebook Search Algorithm Using Principal Component Analysis And Dynamic Subspace Rechecking

摘要

In this paper, we present an adaptive two-stage codebook search algorithm to speed up the encoding process of vector quantization. Initially, the initial codeword in the codebook is determined based on the principal component analysis (PCA) technique. Next, a variant number of its neighboring codewords are rechecked to find out the appropriate codeword in the codebook. The characteristics of the codewords in the codebook are explored to determine the number of search codewords for each input vector. According to the experimental results, it is shown that the proposed algorithm provides a higher image quality while consuming little processing time. In addition, the proposed scheme provides an adaptive mechanism to solve the tradeoff between the encoding time and image quality.