基于新型三层误差反向传播网络的图像压缩  

Image Compression Based Upon New Three Layers Back Propagation

作　　者：刘向阳[1] 王如云[2] 

机构地区：[1]河海大学理学院,南京210098 [2]河海大学交通与海洋工程学院,南京210098

出　　处：《计算机工程与应用》2006年第13期33-35,132,共4页Computer Engineering and Applications

基　　金：国家自然科学基金资助项目(编号:59979004);高等学校博士学科专项科研基金资助项目(编号:1999029410)

摘　　要：基于误差反向传播算法对图像进行压缩的工作已有很多。论文介绍了三层误差反向传播网络拓扑结构和新型二层误差反向传播网络拓扑结构,为进一步提高图像压缩的压缩比和压缩质量,在新型二层误差反向传播网络拓扑结构的压缩层和重建层之间加入一过渡层,使得网络在学习过程中更好地收敛到原始图像。论文提出了新型三层误差反向传播网络拓扑结构并给出了相应算法的严格数学推导过程。经过上机压缩测试,相对于其它误差反向传播算法对图像进行压缩的压缩比、学习速度和压缩质量都有很大提高,取得了很好的效果。There have been many techniques of image compression based upon back propagation arithmetic.In this paper,three layers back propagation networks and the new two layers back propagation networks are introduced.To get the much more quality and rate of image compression,a transitional layer is added in the new two layers back propagation networks.This paper brings forward the new three layers back propagation networks and it＇s arithmetic.It has been proved in many examples that the new networks get a good result in the compression rate,study speed and compression quality of image compression than other back propagations.

关 键 词：图像压缩 人工神经网络 反向传播算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.9[自动化与计算机技术—控制科学与工程]