基于独立分量分析新算法的含噪图像盲分离

New algorithm of blind image separation in noisy mixtures based on independence component analysis

由于乘性噪声的存在,严重限制了标准ICA的使用。在分析独立分量分析的基本模型的基础上,讨论了有噪信号的独立分量分析(Noisy ICA)。提出一种新的基于四阶统计量的方法来消除乘性噪声,分离出独立的源信号。通过寻求噪声线性转换的统计结构,依据代价函数最小来获取解混阵B,从而分离出多维观测信号。最后把算法应用于含噪的混合图像,通过仿真显示算法很好的分离了源信号。

The existence of multiplieative noise greatly limits the applicability of independent component analysis. The basic model of ICA are introduced, and then the ICA of noisy signals is discusse. This paper proposes a method based on fourth-orderstatistic to eliminate multiplicative noise and separate out independent sources. In the paper, the statistical structure of a linear transformation of noisy data is studied, and the statistical structure is used to find the inverse of the mixing matrix by minimization of J. The method is efficient and robust by simulation.