在介绍矢量量化和自组织特征映射神经网(SOFM)的基础上,针对SOFM算法的特点对其进行了几个方面的改进,提高了SOFM网络的性能。采用改进后的基于SOFM的矢量量化技术对图像进行无损压缩编码,码书设计时间减少了约70%,图像效果、编码质量均有所提高,实验结果表明了本算法的压缩比比传统的差值编码(DPCM)无损压缩最高可提升40%,证明了算法的有效性。
The theories of vector quantization(VQ) and self-organizing feature mapping(SOFM) neural networks are introduced in this paper firstly.Some aspects are improved based on VQ of SOFM.The lossless compression algorithm is researched and simulated on the improved SOFM.Codebook design time is reduced by about 70%,and effect of coding image quality is also improved.The compression ration increases 40% in comparison with differential pulse code modulation(DPCM).The results of the experiment illustrate the rationality and efficiency of the algorithm.