一种基于小波域隐马尔可夫模型的SAR相干斑抑制算法

An Algorithm Based on Wavelet-Domain Hidden Markov Models for SAR Speckle Reduction

相干斑噪声是 SAR图像的固有特点。对相干斑抑制的要求是在平滑噪声的同时 ,尽量保持原始图像的结构信息。现有的许多相干斑抑制方法各有优点和不足 ,没有普遍的适用性。基于图像在小波域的隐马尔可夫模型(HMMs)结构 ,结合 SAR图像中相干斑噪声的统计特性 ,本文提出了一种新的小波域相干斑抑制方法。仿真及实测数据处理结果表明 ,该方法在有效抑制相干斑的同时 ,更好地保持了边缘结构。与小波域软阈值去噪方法和 L ee滤波器相比较 ,该方法在噪声平滑及边缘保持上都取得了较大的改进 。

Speckle noise is an intrinsic property of Synthetic Aperture Radar (SAR) imagery. The demand for speckle reduction of SAR images is to smooth the speckle noise while preserving the structure information of the original images. Existing speckle suppression methods possess respective merits and drawbacks, without universal adaptability. Integrating the statistical characteristic of speckle noise in SAR images with wavelet domain hidden Markov models (HMMs) structure of images, we propose a new wavelet domain speckle reduction method. Simulation and experimental results using real data show that the proposed method is able to effectively suppress speckle noise and to better retain edge structure. Compared with wavelet domain soft thresholding denoising algorithm and Lee multiplicative speckle filter, the wavelet domain HMMs method offers significant improvements on smoothing speckle and preserving edge. In addition, the proposed method also gets a better visual effect.