结合加权低秩表示和L_(1)范数的图像混合去噪

Image Hybrid Denoising via Weighted Low-Rank Representation and L1-Norm

由于图像在获取、发布或传输过程中受到噪声的污染,导致图像质量下降.现有的大部分图像去噪方法仅针对高斯噪声情况进行图像恢复,一般来说,现实中图像容易同时受到高斯噪声和椒盐噪声的污染.针对这一情况,文章提出基于加权低秩表示和L_(1)范数的混合噪声去除算法.该算法首先采用加权低秩表示来刻画图像的全局特性,同时利用L_(1)范数来描述稀疏噪声,设计了图像混合去噪模型.然后采用交替方向乘子法对混合去噪模型进行求解.最后对含混合噪声的图像进行了仿真实验分析,结果表明提出的算法能够较好地去除图像中的混合噪声,进一步提高图像的视觉感知质量.

Because the image is polluted by noise in the process of acquiring,publishing or transmitting,the image quality is degraded.Most of the existing image denoising methods only deal with Gaussian noise for image recovery.Generally speaking,real image is easily affected by both Gaussian noise and salt and pepper noise.To address this problem,a hybrid noise removal algorithm based on weighted low rank representation and L1 norm is proposed.Firstly,the weighted low-rank representation is used to describe the global characteristics of the image,and the L1 norm is used to describe the sparse noise.Then the mixed denoising model is solved by using the alternating direction multiplier method(ADMM).Finally,the simulation results show that the proposed algorithm can remove the mixed noise and improve the visual quality of the image.