基于多子波的SAR图像噪声的抑制

SAR图像的噪声为乘性噪声，传统的图像去噪方法效果较差。SAR图像噪声抑制的方法一般可分为空域滤波和频域滤波。子波分析是一种典型的频域处理方法，通常首先对SAR图像进行子波分解，保留低频，对高频子带的系数做硬阈值或软阈值处理，然后进行重构。对于平滑区域，噪声抑制效果明显，但在边缘细节比较丰富的区域，细节损失严重。本文在多子波的预滤波中采用了冗余技术，并利用边缘跟踪算法对子波分解后的高频子带中的边缘和噪声进行有效的分离，从而确保了在抑制噪声的同时保留了边缘信息。实验结果验证了该方法的有效性。

Suppression of SAR Image Speckle Based on Multiwavelet

As a classic method in frequency domain, wavelet analysis begin with decomposing the SAR image, then get the low frequency coefficients which will be kept and soft or hard threshold is applied on the high ones, at last all of them are reconstructed. The noise in the smooth area can be clearly suppressed, but in the areas with rich detail information edge the information detail is lost. The paper presents a method based on multiwavelet. Repeat-row approach is applied in the prefilter of multiwavelet to improve its denoising performance. At the same time the edge and noise which are hidden in the high frequency coefficients are separated by edge tracking algorithm, so that edge information is kept and the noise is suppressed. The numerical examples have proved that the method is more valid than others.