一种用于图象编码的神经网络及其改进算法

A Neural Network and Its Improving Algorithms Used in Image Coding

本文用Kohonen的自组织特征映射神经网络设计图象矢量量化的码书,研究了网络的基本性质和学习算法的实现,提出了对学习算法的改进方法。实验结果表明,自组织特征映射神经网络能够有效地用于构造图象矢量量化的码书,算法简洁,实现快速,编码效果优良。

The kohonen 's self-organizing feature map neural network is used to design codebooks of image vector quantization in this paper . The basic properties of the network and the realization of the learning algorithm are studied , and the improving methods of the learning algorithm are introduced . The experimental re- sults show that the self-organizing feature map neural network can be effectively used to construct codebooks of image vector quantization . The algorithms are simple , the realizations are quick , and the coding results are good.