基于自适应过完备稀疏表示的红外图像滤波方法

Infrared Image Denoising Based on Adaptive Over-complete Sparse Representation

摘要:

针对传统滤波算法在滤除红外图像噪声时会损失部分有用信息的问题,提出一种基于自适应过完备稀疏表示的红外图像滤波方法。该方法采用K-SVD算法以待滤波的红外图像为样本训练出自适应过完备原子库;采用正交匹配跟踪算法将红外图像信号在该过完备原子库上稀疏分解为稀疏成分和其他成分,稀疏成分对应红外图像中的有用信息,其他成分对应红外图像中的噪声,由稀疏成分重建图像,从而达到消除噪声的目的。实验结果表明:该方法相比传统方法具有更好的滤波效果,重建图像质量较高。 更多还原

Abstract:

As traditional infrared image denoising methods usually achieve the satisfactory effect at the cost of damaging the image content, an infrared image denoising method based on adaptive over-complete sparse representation is presented. The image can be decomposed via orthogonal matching pursuit algorithm into the content part and the remainer on the adaptive dictionary trained by the K-SVD method in allusion to the infrared image, eventually the infrared image can be reconstructed by the content part. Experimental results show the effectiveness of the method.