基于贝叶斯估计的改进Contourlet变换的SAR图像滤波

针对小波变换和轮廓波变换在合成孔径雷达(Synthetic aperture radar, SAR)图像中去噪应用的不足,结合小波变换和轮廓波变换的优点,将小波变换与轮廓波变换相结合,提出一种改进轮廓波变换方法。首先将待处理图像进行小波变换,然后对低频子带图像进行重建,得到一个细节子带图像,然后使用方向滤波器组对其进行多方向划分。再采用贝叶斯最大后验概率估计对划分后的方向子带信号进行估计。实验结果表明此方法在抑制图像斑点噪声的同时,很好的保持了均匀区域的辐射特性,保持了图像中的边缘以及细小纹理,且没有人为畸变。此外算法的高频子带图像含有更为丰富的纹理,对于边缘特征的提取非常有益。

Improved Denoising Method of Contourlet of SARImage Based on Bayesian Estimation

According to shortcoming of Wavelet and Contourlet in the applicatiopn of SAR image denoising, taking advantage of wavelet and contourlet, the proposed method was complemented by wavelet transform in conjunction with directional filter banks. First, the proposed method derived wavelet coefficients of entry image by wavelet transform, then reconstructed the image only by lowpass coefficients and after processing it can derived a highpass subband, afterwards to partition the highpass subband into multi-direction subbands by direction filter banks. Last we adapt Bayesian maximum aposteriori probability(MAP) estimation to estimate these multi-direction subbands. The experimental results show that the proposed method can restrain the speckle noise effectively at the same time it can keep the Radiated characteristics of even region very well, it also can keep edge information effectively and avoid artifacts. Futhermore the highpass subband image of the proposed method contains more texture, it is beneficial to extract edge feature of image.