基于剪切波和低秩稀疏表示的噪声图像融合算法研究

Noisy image fusion algorithm based on shearlet and low-rank sparse representation

多源图像在获取和传输过程中不可避免地会引入一些噪声,如何利用有效的数学工具和方法设计具有抗噪声的图像融合算法特别受到关注.提出了一种基于剪切波多尺度变换和低秩稀疏表示的图像融合去噪算法.首先对两幅多聚焦噪声图像进行剪切波变换,对变换后获得的低频子带采用基于区域能量匹配度的融合规则;而对所获得的高频方向子带则采用低秩稀疏表示的融合规则,即分别通过低秩准则和稀疏准则来捕捉高频子带的整体结构和局部结构信息,同时对经过低秩稀疏分解所获得的噪声矩阵进行去噪.大量实验表明,所提出的方法可有效实现多聚焦图像的融合,更好的保留了图像的细节纹理信息,同时能够很好地对源图像中的噪声进行去除,使融合后的图像更加清晰.

Some noises will be introduced in the process of acquisition and transmission of multi-source images inevitably,so how to effectively use mathematical tools and methods to design image fusion algorithms with anti-noise is of increasingly critical concern.This paper proposes an image fusion denoising algorithm based on shearlet multi-scale transform and low-rank sparse representation.The algorithm first performs shearlet transformation on two multi-focus noise images.The regional energy matching fusion rule is used for the low-frequency subbands,while the low-rank sparse representation fusion rule is used for the high-frequency direction subbands.Meanwhile,the noise matrix obtained by low-rank sparse decomposition is denoised.Sufficient experiments show that the proposed method can effectively achieve the fusion result of multi-focus images,better retain the detailed texture information of the image,and can also remove the noise in the source images well and achieve better fusion performance.