| 基于自组织特征映射的矢量量化方法 |
| |
| 引用本文： | 林昌,康泰兆. 基于自组织特征映射的矢量量化方法[J]. 南京理工大学学报(自然科学版), 1999, 23(5): 393-396 |
| |
| 作者姓名： | 林昌 康泰兆 |
| |
| 作者单位： | 1. 福州电信局,福州,350005
2. 南京理工大学信息学院,南京,210094 |
| |
| 摘 要： | 对自组织特征映射神经网络的特性进行分析，并将其与矢量量化问题的实质进行比较，提出了一个实现矢量量化的自组织特征映射算法。分析与实验表明，该算法是稳定收敛的。算法的学习结果与网络的初始状态无关，并且十分接近于全局最优解的下限。将该算法应用于图像数据的压缩，取得了很好的结果
|
| 关 键 词： | 图像处理 神经网络 特征抽取 |
| 修稿时间： | 1997－08－30 |
A SOFM Algorithm for Vector Quantizing |
| |
| Lin Chang,Kang Taizhao. A SOFM Algorithm for Vector Quantizing[J]. Journal of Nanjing University of Science and Technology(Nature Science), 1999, 23(5): 393-396 |
| |
| Authors: | Lin Chang Kang Taizhao |
| |
| Affiliation: | (Fuzhou Telecom,Fuzhou 350005);( |
| |
| Abstract: | The characteristics of SOFM neural network is analysed and compared with the feature of Vector Quantizing problem in this paper. Based on this an algorithm for Vector Quantizing is put forward. Analysis and experiments show that this algorithm is stably convergent. The study results are irrelevant to the initial status of the network and quite approximate to the lower limit of the optimum solutions. Very satisfied results are achieved when this algorithm is used for graphic data compression. |
| |
| Keywords: | image processing neural networks feature extraction |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
