Codebook design by a hybridization of ant colony with improved LBG algorithm

Ant colony algorithm is a newly emerged stochastic searching optimization algorithm in recent years. In this paper, an appropriately adapted ant colony system embedded with a simple improved LBG algorithm is proposed for vector quantization codebook design. The emphasis is put on the design of the probability transfer function and the tabu list in the ant colony algorithm, the utilization of the next nearest neighborhood in the LBG algorithm, as well as the update of the pheromone in both local and global sense. Experimental results show that the new algorithm outperforms other well-known codebook design algorithms, and particularly, the improvement of PSNR exceeds 2 dB compared with the conventional LBG algorithm.