摘要
针对SAR图像相干斑的随机性,采用统计信号处理中盲源分离的方法进行去噪。方法利用了基于峰度的盲源分离(BSS)开关算法,从轮廓(contourlet)变换域和小波的多方向滤波器分解(DFB)的变换域中,进行盲源分离以达到SAR图像的去噪。实验结果显示,与contourlet变换域的其他去噪方法相比,基于峰度的BSS开关算法不仅从视觉上对图像的质量有明显改进,而且在量化图像指标上也得到了较大的提高;而contourlet变换域与小波的多方向分解(DFB)变换域的相应的各类去噪方法相比,后者能得到更好的结果。
In view of the random nature of speckle noise of SAR image, a method based on blind source separation used in statistical signal processing was proposed for wiping off the speckle noise of SAR image in this paper. This method took the kurtosis as the optimizing function of the switching algorithm of blind source separation. In order to get rid of the speckle noise of SAR image, the image was translated into the contourlet area and ameliorated wavelet area, then separated the coefficients of translation area. The experimental results show that the results using this methods are better than that using the other methods of translation area. And the new translation area has preferable results. It materializes that not only a good vision of image was gotten but the quality indices were greatly improved.
出处
《计算机仿真》
CSCD
北大核心
2009年第9期164-167,共4页
Computer Simulation
基金
国家自然科学基金(60375003)
航空基础科学基金(03I53059)
关键词
轮廓变换
小波
方向滤波器组
隐马尔科夫树模型
盲源分离
Contourlet transformation
Wavelet
Directional filter banks （ DFB ）
Hidden Markov tree （HMT）
Blind source separation （BSS）
