中文摘要：

为了实现大压缩比的遥感图像压缩，利用神经网络的自组织、并行计算和分布式存储的能力，提出一种基于神经网络的压缩方法．在传统单隐层前向神经网络的基础上，该网络使用一种新的能有效处理直线型和曲线型奇异性的多尺度几何分析工具．脊波，作为隐层神经元的激活函数．它不仅具有神经网络压缩的优点；并且由于脊波良好的时、频和方向局域化特性，能够对遥感图像的边缘和轮廓实现更加有效的表示．仿真结果表明：该方法不仅能实现较高的压缩比，而且具有重建图像质量好、学习快速和鲁棒性强等优点．

英文摘要：

To get a high-ratio compression of remote sensing images, a neural network （NN）-based compression method was advanced. By using the characteristics of self-learning, parallel processing and distributed storage of NN, a single hidden layer feed-forward NN was constructed for getting high-ratio compression of remote sensing images. Moreover, we employ fidgelet, which is a new geometrical muhiscale analysis （GMA） tool and is powerful in dealing with linear singularities （and curvilinear singularities with a localized version）, as the activation function in the hidden layer of the network. Therefore the network has both the advantages of NN-based image compression method and more effective representation of edges and contours for the localization properties of ridgelet in scale, location and direction. The simulation results show that the proposed network can not only get high compression ratio but also present promising results, such as high reconstruction quality, fast learning and robustness, as compared to available techniques in the literature.