基于阻尼修正FICA的图像分离算法

An Image Separation Algorithm Based on Damping Modification FICA

独立分量分析算法是一种常用的盲源信号(包括图像信号、音频信号等)分离方法,它的主要任务在于分析混合矩阵及对应的分离矩阵的结构。文章提出了一种阻尼修正的独立分量分析方法,它可以根据迭代过程中梯度函数的收敛情况来动态设置和调整阻尼系数的值,从而保证梯度函数始终能收敛于某一稳定值,并且最终可提取出一组特定的独立分量,同时还具有修正的独立分量分析方法所具有的计算速度快的优点。通过两组混合图像的分离实验,证明该方法具有一定的实用意义。

An independent component analysis algorithm, usually used to separate bind source signs, such as audio and images, etc., whose main goal is to analyze the structures of a mixed matrix and its separation matrix. This paper proposes an independent com-ponent analysis algorithm of damping modification, which may adjust the damping coefficient dynamically according to the conver-gent state of gradient function during iteration, make sure that gradient function converges to a certain stable value. And a set of partic-ular independent components can be drawn finally. Otherwise, it has an advantage of converging fast. An experiment of separating a set of mixing images proves that the methodology has some practical meanings.