一种新的SAR图像相干斑抑制算法(英文)

A new method for SAR images speckle reduction

合成孔径雷达(SAR)图像会受到相干斑噪声的污染,对SAR图像的后续处理产生了很大影响。提出一种基于快速离散曲波变换(FDCT)抑制合成孔径雷达(SAR)图像相干斑噪声的方法。先通过FDCT把SAR图像变换到曲波域中,得到曲波系数,再应用自适应阈值算法估计不同尺度、不同方位曲波系数的阈值,分别对曲波系数进行硬阈值和软阈值化处理,最后通过FDCT反变换恢复出图像。对单视SAR原始图像进行处理,并与小波去噪方法进行各种量化比较,结果表明,Curvelet滤波器要比Wavelet滤波器效果好,软阈值算法的效果比硬阈值算法好。基于FDCT的SAR图像相干斑去噪,不仅抑制相干斑能力比较强,而且在目标的边缘及纹理信息的保持上也有很大的优势。

Synthetic Aperture Radar （SAR） image is polluted easily by speckle noise, which can affect further processing of SAR image. Traditional methods employ wavelet transform, which is only effective in representing point singularities. Based on Fast Discrete Curvelet Transform （FDCT）, a de-noising method for SAR image is presented. FDCT is employed to transform the SAR image into the curvelet domain to obtain the curvelet coefficients, and then soft and hard thresholding de-noising processes are performed separately on the Curvelet coefficients of different scales and directions by using adaptive threshold estimation. Finally the SAR image is reconstructed by inverse FDCT. This de-noising method is applied to the experiments of a single look SAR image, and compared with the wavelet de-noising method. Experimental results indicate that based-FDCT de-noising method is a more effective method, which is not only better in reducing speckle, but also of advantage in holding information of target edge and grain.