基于Shearlet域系数处理的SAR图像降噪

发布时间：2018-09-01 07:49

【摘要】：结合图像在Shearlet域中系数的特点,提出了一种基于Shearlet系数稀疏表示与投影总变分(total variation,TV)相结合的合成孔径雷达(synthetic aperture radar,SAR)图像去噪算法。有效解决了稀疏表示在图像去噪时存在的边缘细节损失与TV去噪时存在的光滑区域阶梯效应。首先,利用SAR图像Shearlet系数的稀疏性,结合系数稀疏表示模型,采用分段正交匹配追踪方法求解优化解,从统计意义上实现稀疏表示后的系数均值为真实图像系数均值的无偏估计;其次,为弥补稀疏表示中丢失部分系数在图像细节上的损失,同时结合这部分系数对应的Shearlet函数有利于表征图像边缘细节的特性,针对图像在丢失系数对应的Shearlet函数空间中投影重构的结果,结合TV方法迭代去噪。实验结果表明,该方法充分利用Shearlet域系数的特性,采用稀疏去噪与投影TV相结合的方法以弥补各自缺陷,在去噪的同时能有效保持图像纹理细节,并具有更优的图像视觉效果。
[Abstract]:A new synthetic aperture radar (SAR) image denoising algorithm based on the combination of Shearlet coefficients sparse representation and total variation (TV) is proposed, which effectively solves the edge detail loss of sparse representation in image denoising and the existence of TV denoising. Firstly, by using the sparsity of the Shearlet coefficients of SAR images and the sparse representation model of the coefficients, the optimal solution is solved by piecewise orthogonal matching pursuit method, and the unbiased estimation of the sparse representation coefficients to the real image coefficients is realized statistically. Secondly, the missing coefficients in the sparse representation are compensated. The loss of some coefficients in image details and the Shearlet function corresponding to these coefficients are helpful to characterize the edge details of the image. According to the result of projection reconstruction of the image in the Shearlet function space corresponding to the loss coefficients, the method is combined with TV iterative denoising. The experimental results show that this method makes full use of the coefficients of Shearlet domain. The method of combining sparse denoising with projection TV is used to compensate for their defects. The method can effectively preserve texture details while denoising, and has better image visual effect.
【作者单位】：重庆大学通信工程学院;
【基金】：国家自然科学基金(61301224) 重庆市自然科学基金(cstc2012jjA40001) 中央高校基本科研业务费(CDJZR11160003,CDJZR12160014)资助课题
【分类号】：TN957.52

【参考文献】