基于小波变换与组稀疏相结合的遥感图像复原算法

Remote Sensing Image Restoration Method Based on Wavelet Transform Combined with Group Sparse

针对脉冲噪声下的遥感图像的复原问题,本文设计一种基于小波变换与组稀疏相结合的遥感图像复原算法。该算法模型采用L0范数作为数据保真项,可以有效地去除脉冲噪声;在正则项上,本文采用重叠组稀疏正则化器与梯度图像在小波变换下的L0范数进行稀疏建模。使用优化最小化方法分别与交替方向乘子法对算法模型进行求解。将本文复原算法与L0-OGSTV、HNHOTV-OGS、L0-TV三种算法进行实验对比。实验结果表明,在峰值信噪比和结构相似性的指标上,本文算法均优于以上几种算法。

To solve the restoration problem of remote sensing images under impulse noise, an image restora-tion algorithm based on wavelet transform combined with Group sparse is designed in this paper. The proposed algorithm utilizes L0 norm as the data fidelity term, providing an effective means for removing pulse noise. In the regularization term, a L0 norm of gradient images under wavelet transform is implemented and an overlap-group sparsity regularizer for sparse modeling. The algo-rithm is solved through optimization minimization methods and alternating direction of multiplier approach, respectively. The restoration algorithm in this paper is compared with L0-OGSTV, HNHOTV-OGS and L0-TV. The experimental results show that this algorithm is superior to the above algorithms in terms of peak signal-to-noise ratio (PSNR) and structure similarity (SSIM).