加权平方误差测度下的快速矢量量化算法

为降低加权平方误差测度下的矢量量化运算量,针对加权因子固定与不固定2种情况,分别提出了快速搜索算法。加权因子固定时,对等均值最近临搜索算法做了相应改动即可应用;加权因子随输入矢量变化时,提出了一种分裂多级等均值最近临搜索算法,算法提出了3个新的排除准则,在不同的场合下选用部分或者全部,从而有效降低码字搜索运算量。测试结果表明:分裂多级等均值最近临搜索算法能够有效降低加权平方误差测度下矢量量化的运算量,比全搜索算法能够节省约69%的运算量。

Fast search algorithm for vector quantization based on the weighted squared error

Fast search algorithms were developed to reduce the computational load for vector quantization based on the weighted squared error,static weighted coefficients,and dynamic weighted coefficients.The equal-average nearest neighbor search algorithm(ENNS) was modified for applications with static weighted coefficients.A split multi-stage ENNS(SMSENNS) algorithm was developed for applications with weighted coefficients that change with the input vector.The SMSENNS algorithm was three new rejection principles for different conditions to reduce the search load.Test results show that the SMSENNS algorithm efficiently reduces the computational load by nearly 69% compared to the full search algorithm.