论文部分内容阅读
给出了一种基于双尺度字典的极化合成孔径雷达(Pol SAR)图像滤波算法。首先,将原始极化SAR图像进行区域分类;然后,对不同匀质区域数据采用不同尺度字典,并进行字典训练和稀疏去噪;最后,重构去噪结果,形成去噪后的SAR图像。采用美国AIRSAR实测半月湾数据进行实验,并与单尺度稀疏去噪算法的结果进行比较,结果表明本算法的斑点噪声抑制能力更强,边缘纹理细节更加清晰,此外,在强散射点的点目标幅值特征保持和SAR图像极化特性保持两个方面也取得很好的效果。
A polarimetric synthetic aperture radar (Pol SAR) image filtering algorithm based on double-scale dictionary is presented. Firstly, the original polarimetric SAR images are classified by region. Then, different scales of dictionaries are used for different homogeneous regions, and dictionary training and sparse denoising are performed. Finally, the denoised results are reconstructed to form denoised SAR images. The experimental data of half moon bay measured by AIRSAR in the United States were compared and compared with the results of single-scale sparse denoising algorithm. The results show that the proposed algorithm has stronger speckle noise suppression and clearer edge texture details. In addition, The preservation of amplitude characteristics and the polarization characteristics of SAR images also achieved good results in two aspects.