中文摘要：

提出一种基于四树复小波包变换的局部窗口阈值SAR图像去噪新方法。该方法利用四树复小波包变换具有的移不变性、良好的方向选择性和对高频信号的细致分析能力等特点，把含噪SAR图像分解成低频逼近子图和若干高频方向子图。通过对方向子图设置合理的阈值来确定最优复小波包基。在保留低频逼近子图复系数不变的同时，利用高频信号系数的邻域相关性和噪声方差随分解尺度增大而迅速衰减的特点，对最优基复小波包系数进行局部邻域窗口阈值收缩处理，从而实现降噪功能。实验结果表明，该方法计算效率高，在等视指数（ENL）、优点图（FOM）等指标上均优于传统的复小波变换、复小波包变换和Curvelet域HMT等去噪方法，能有效地抑制SAR图像斑点噪声的同时，对图像边缘和细节具有较好的保护能力。

英文摘要：

A novel SAR image denoising scheme based on quad-tree complex wavelet packet transform （QCWPT） is presented using local neighborhood window threshold method. The noisy SAR image is decomposed into a low frequency approximation sub-image and some high frequency directional sub-images via the QCWPT, which has shift invariance, good directional analysis ability and ability to analyze the high frequency detail signal carefully. The best complex wavelet packet basis is determined via setting reasonable threshold in high frequency directional sub-images. The complex coefficients in the low frequency approach sub-image remain unchanged, and according to the neighbor-hood correlation of the high frequency signal coefficients and the characteristic that the variance of the noise attenuates rapidly along with the increasing of the decomposition scale, a local neighborhood window threshold method is used for shrinking high frequency complex coefficients in the best complex packet basis. Experiment results indicate that the presented scheme has higher operational efficiency and outperforms the traditional dual-tree complex wavelet transform （DCWT）, QCWPT and Curvelet domain Hidden Markov Tree in terms of the equivalent number of looks （ENL）, the figure of merit （FOM） and visual effects. Experiments also show that the presented scheme could not only remove noises effectively, but also reserve the texture and edge details of the SAR image better.