图象矢量量化器设计的研究
An Approach to the Design of the Vector Quantizer for Image Signals
-
摘要: 矢量量化是近年用于图象数据压缩的一种高性能的方法,它把信号作为矢量而不是作为标量来编码。在矢量量化器的设计过程中,过程的初始化对于整个设计的计算代价(收敛速度)以及所达到的指标(平均失真)有重大的影响。本文提出了一种"误差函数初始化"(EFI)的方法,这种方法直接针对失真准则,通过求"误差降低函数"的估计量的最大值而使"种子"的选择最佳化。实验结果以及在此基础上的分析表明了EFI作为初始化方法的优点,并且显示了其作为实时矢量量化器刷新算法的前景。Abstract: In image data compression vector quantization has demostrated noticeable performance.The design method for an optimized vector quantizer, however, is rather an expensiveprocess. The paper presents a technique, which, besides being able to improve theoptimizing design algorithm in the design cost and the performance index, can also serveas a fast design method for sub-optimal vector quantizers in real-time systems.