摘要
讨论了超定情况下基于稀疏成分分析的盲图像源分离的问题,对线性聚类的条件及超定情况基于线性聚类的盲源分离做出了证明,通过准确的估计混合源的数目,对小波域基于稀疏成分分析的分离算法做了改进,给出了一种超定情况下的盲图像源分离方案.实验结果表明,改进的盲源分离算法能实现超定情况下的盲图像源分离,而直接利用稀疏成分分析的方法及经典的FastICA的方法分离结果均欠佳.
In this paper, the problem of over-determined blind image source separation based on sparse component analysis was discussed, and the conditions of linear cluster and over-determined blind source separation were proved. The separation algorithm based on sparse component analysis in wavelet domain was improved by estimating the numbers of sources. Experiment result shows that improved algorithm can extract all the image sources under over-determined conditions, while the sparse component analysis method used directly and the classical FastlCA algorithm cannot separate the mixtures perfectly.
出处
《曲阜师范大学学报（自然科学版）》
CAS
2012年第3期61-64,共4页
Journal of Qufu Normal University(Natural Science)
基金
山东省优秀中青年科学家科研奖励基金(BS2010DX012)
曲阜师范大学基金(XJ201010)
关键词
超定
稀疏成分分析
盲源分离
over-determined
sparse component analysis
blind source separation
