一种非零元个数约束的字典学习图像去噪算法*

提出了一种基于字典学习的图像去噪算法。在K-SVD字典学习算法的基础上，改变稀疏编码中误差约束为非零元个数约束来进行字典学习。在实验的基础上分析了使用不同非零元个数去噪时对峰值信噪比的影响，提出分别针对低噪图像和高噪图像采用两个固定非零元个数来进行字典学习，获得图像的稀疏表示，从而恢复出原始图像。实验结果表明，与小波软阈值去噪方法相比，本算法能够在保留图像边缘和细节信息的同时有效地去除图像中的噪声，具有较好的视觉效果。

Image denoising algorithm of dictionary learning restricted by nonzeros number

This paper presented a image de-noising algorithm based on dictionary learning.It implemented dictionary learning by substituting error restriction for non-zero number restriction on the base of K-SVD method.By analyzing effection on peak signal-to-noise ratio of denoising by different nonzeros number,it proposed implementing dictionary learning by using different number non-zero respectively for low-noise image and high-noise image to obtain sparsity represetation of image and restore original image.The re...