基于小波变换和ICA的新型有噪混合图像盲分离方法

Separation method of noisy mixed images based on wavelet transform and ICA

独立分量分析虽能有效地对无噪信号实现分离,但是直接应用于有噪信号时效果较差.论文针对这个问题,给出了一个消噪-分离-消噪策略,并将之用于有噪混合图像盲分离且给出了具体的分离方案.在该方案中,首先利用小波变换对有噪图像进行消噪,然后再使用独立分量分析对消噪后的图像进行分离,最后再一次利用小波变换对分离后的图像再次消噪,从而获得较为清晰的图像.仿真实验表明,该方法能有效提高有噪混合图像分离结果的峰值信噪比和相关系数,效果良好.

Independent component analysis （ICA） can effectively separate noiseless signal,but can not effectively be applied to noisy signal.To solve this problem,a separation method of noise signal based on denoising-separation-denoising strategy was proposed and was applied to separate noisy mixed images,and a specific separation scheme was given.The presented method firstly employs wavelet transform denoising,and then uses ICA separate denoised image,finally applies wavelet transform denoising again.Simulation experiments show that this method can effectively improve signal-to-noise ratio and correlation coefficient on peak of separation of noisy mixed images.