神经网络在图像压缩中的应用研究

Research of neural network application in image compression

探讨应用神经网络实现图像压缩和解压缩技术。神经网络结构由两组三层ＢＰ网络嵌套而成，分别为输入层、结合层、压缩层、分解层和输出层。由ＲＧＢ彩色模式的ＢＭＰ图像文件产生样本数据，采用ＢＰ算法的一种扩展型，即嵌套训练算法，对网络进行训练，得到有效的模拟结果。

Using of neural network in image compression and decompression discussed. The neural network consists of two nested three-layers BP networks. They are input layer, combination layer, compression layer, decompression layer and output layer respectively. The extension of BP Algorithm-NTA (Nested Training Algorithm) is adopted in training of it. The simulation results indicate its efficiency.