To reduce speckle noise and preserve edge characteristics in synthetic aperture radar (SAR) images, an additive transform noise mode of speckle noise in the SAR image is given and a new algorithm for speckle reduction by the combination of self-snake diffusion and regulated L1-L2 optimization under undecimated wavelet packet transform (uWPT) is proposed. In the new method, a SAR image is first decomposed into multiple subbands by multi-level uWPT. The lowpass subband is filtered by self-snake diffusion, and the subband filtered is regarded as the local mean of the original SAR image in the wavelet domain. Based on the local mean, the adaptive and shrinkage soft-thresholding filter is applied to the remaining subbands by regulated L1-L2 optimization. Finally, the despeckled image is recovered from all of filtered subbands by the inverse uWPT. Experimental results show that compared with the Kuan filter algorithm, the P-M diffusion filter algorithm and the Γ-WMAP algorithm using undecimated wavelet transform, the proposed algorithm has better performance in terms of reducing speckle noise and preserving the edge of SAR images.%为抑制合成孔径雷达(SAR)图像乘性相干斑噪声,同时有效保护SAR图像的边缘特征,给出了相干斑噪声在小波域的一种加性转换噪声模型.并以此模型为基础,提出了一种非下采样小波包分解下自蛇扩散与改进I1-L2联合优化相结合的相干斑噪声抑制新算法.该算法利用非下采样小波包变换对SAR图像进行多层子带分解,然后对低通子带系数进行自蛇扩散滤波,并将滤波处理后的系数作为原SAR图像在小波域的局部均值估计,再以此局部均值为基础,利用改进的L1-L2联合优化对其他各高频子带系数进行自适应软阈值收缩滤波去噪.最后通过重构滤波后的各子带系数实现SAR图像相干斑噪声抑制.实验表明:与经典的空域Kuan滤波算法、P-M扩散滤波算法及基于非下采样小波变换的Γ-WMAP算法相比,本算法在SAR图像的相干斑噪声抑制与边缘保护方面均取得了较好的效果.
展开▼
