神经网络—空间方向小波四叉树压缩编码

Neural network-spatial orientation wavelet quad-tree imagecompression coding

将空间方向小波四叉树编码与自组织特征映射神经网络相结合,提出了一种新的多尺度系数矢量量化压缩策略。首先通过小波分解得到三个方向的高频多尺度系数矢量,分别利用自组织特征映射神经网络对三个方向的多尺度系数矢量进行加权矢量量化压缩编码。仿真实验结果表明本文提出的算法是合理可行的。

Based on spatial orientation wavelet quad-tree and self-organizing neural network, a new vector quantization image compression coding method is presented. Firstly three directional high frequent multi-scale coefficients vector is formed by wavelet decomposition, then self-organizing neural network is used to compress coding. Simulation results show that the algorithm presented is feasible.