采用外部数据扩充样本的协同表示去噪算法

Cooperative representation denoising algorithm using external data to expand samples

针对滤波算法利用内部数据很快达到滤波性能极限、在噪声密度较大时无法满足滤波要求的问题,提出采用外部数据扩充样本的协同表示去噪算法。通过协同表示寻找到与待滤波图像相似的外部数据图像,并利用该图像生成虚拟图像,扩充数据库纹理信息;通过匹配待滤波图像与外部数据图像的相似碎片信息,完成滤波过程。实验结果表明,所提出算法对高密度噪声图像的滤波效果较BM3D算法有大幅度的提升,并具有较好的鲁棒性,可获得高质量的滤波图像。

In view of the fact that the filtering algorithm using internal data reached the limit quickly and can not meet the requirements when the noise density was large,a cooperative denoising algorithm was proposed based on external data expansion samples.The external data image which similar to the image to be filtered was found through the cooperative representation,and the virtual image was generated from the image to expand the texture information of the database.The filtering process was completed by matching the similar fragmentation information between the image to be filtered and the external data image.The experimental results show that the proposed algorithm can greatly improve the filtering effectness of high-density noise images compared with the BM3D algorithm,and obtain high-quality filtered images with good robustness.