Fast Codeword Search Algorithm for Split-dimension Vector Quantization based on the Sequence of Characteristic Value

A Fast Codeword Search Algorithm for Split-dimension Vector Quantization based on the Sequence of Characteristic Value is proposed. Firstly, an improved PSO (Particle Swarm Optimization)method is adopted to design split-dimension vector codebook in LCT (Local Cosine Transform) domain. Before coding, for each splitted vector( a vector splitted into a few sub-vectors), the characteristic value of each codeword is calculated and ranked. During coding, the search sequence is decided by the characteristic value of each input vector. In the meantime, the search range is confined and an efficient elimination criteria is used in order to accelerate the coding process significantly. A great deal of experimental results shows that this algorithm can achieve a similar performance to full search algorithm while the coding time is only 1.7%∼5.0% of full search process.