基于自相关函数建模的小波域SAR图像去斑方法

Speckle Reduction of SAR Image Based on Modeling of Autocorrelation Function of Wavelet Coefficients

合成孔径雷达(SAR)图像的小波系数间存在重要的相关性,通过对这种相关性的精确建模可以改善图像的去斑效果。提出了一种新的基于自相关函数建模的小波域SAR图像去斑方法。首先对原始SAR图像进行对数变换,再用可控金字塔作多尺度和多方向分解,分别对图像和噪声系数的自相关函数精确建模,并在图像自相关函数中引入方向性解析式,再利用维纳滤波得到去噪后的小波对数图像,最后经指数变换得到去斑后的SAR图像。对合成图像和实际SAR图像的去斑实验表明,该方法较其他经典方法的去斑效果要好。

The wavelet coefficients of SAR images have important correlations, and correct modeling of the correlations can improve despeckling effect. A new method to suppress speckle noise of SAR images is proposed in this paper. The main contributions of the method are to accurately model the autocorrelation function of wavelet coefficient of logarithmically transformed SAR images in steerable pyramid and to incorporate the directionalities of subband coefficients into the model in an analytic way. Based on the correlation models of signal and noise. Wiener filter is applied to separate speckle noise from original SAR images. The simulation results on simulated images and real SAR images show that the proposed method has better performance than the previous methods.