面向输电线路的压缩感知图像去噪方法

传统的基于字典学习的输电线路图像去噪方法，易受冗余字典影响存在重建图像边缘细节恢复不足的问题.为了有效抑制输电线路图像表面存在的高斯噪声，提出一种图像非局部自相似特性与改进K-SVD字典学习算法融合的输电线路图像去噪方法，利用图像非局部自相似性作为正则项约束并加权稀疏表达模型，提高去噪图像复原和保留细节的能力.实验选取含有自然图像和输电线路典型缺陷图像进行仿真实验测试.实验结果表明，所提出的算法不仅能够很好的保留图像纹理特征与边缘细节，对高斯噪声也具有良好的鲁棒性.

Compressed sensing image denoising method for transmission lines

The traditional dictionary-based signal line denoising method for transmission lines is vulnerable to the redundancy dictionary and causes insufficient restoration of the edge details of reconstructed images. In order to filter out the Gaussian noise existing on the surface of transmission line image effectively， an image denoising method combining non-local self-similarity of image and K-SVD (K-means and Singular Value Decomposition) dictionary learning algorithm is proposed. Similarity is used as a regular term constraint and weighted processing to improve the quality of denoising image restoration. The experiment selects several typical defects (broken strands， wear， bubbles) of the transmission line for simulation test. The experimental results show that the proposed algorithm can not only preserve the image texture features and edge details， but also has good robustness to Gaussian noise.