Fast image artistic style learning using twin-codebook vector quantization

Abstract

Recently, the Twin-codebook Vector Quantization (TBVQ) has been developed for image colourization. It adopts two different but related codebooks in the encoder and decoder respectively, instead of the same codebooks used in traditional vector quantizers. In this paper, a novel image artistic style learning scheme is proposed based on TBVQ. During the learning process, the target image is first encoded by the natural codebook and then decoded by the stylized codebook. Effectiveness of the proposed method is shown by experimental results. © 2012 ISSN 2073-4212.