利用二阶统计量的图像盲提取方法

Bind image extraction method using second-order statistics

为了从混叠的图像信号中分离源图像,提出一种基于二阶统计量的图像盲抽取方法。首先根据图像信号的非平稳特性构造用于估计分离向量的代价函数;然后通过最小化其函数获得最优分离向量,并实现逐次恢复源图像信号的目的。仿真实验结果表明,本文方法不仅能较好地实现多幅图像的盲分离,还能同时分离服从亚高斯分布的图像信号和超高斯分布的语音信号;与其它传统方法相比较,它具有更优越的估计性能。

To separate the source image one by one,we consider the blind image extraction of instantaneous mixtures using second-order statistics.A new cost function was constructed first by exploiting the non-stationary properties of image signals,and then the optimal extracted vectors were determined through minimizing the cost function,so as to separate the source image one by one.The simulation results show that the method can achieve the blind separation for mixed images.Moreover,it can separate images with sub-Gaussian distribution and speeches with super-Gaussian distribution.Compared with other conventional algorithms,it has higher accuracy.