An efficient blind image separation based on screening single source points in transform domain

作者：Xu Jin Dong; Yu Xian Chuan^*; Hu Dan; Zhang Li Bao

来源：Acta Electronica Sinica, 2013, 41(11): 2193-2198.

DOI：10.3969/j.issn.0372-2112.2013.11.013

• Blind image separation
• Latent variable
• Separation efficiency
• Signal features
• Single source
• Sparse component analysis
• Statistical histograms
• Transform domain

In the field of blind image separation based on sparse component analysis, the separation efficiency and accuracy is directly affected by the valid number of clustering samples. For this problem, a new algorithm for detection of points in the Haar wavelet domain where only single source contributions occur was proposed. The algorithm identified the single source points (SSPs) by comparing the absolute direction between diagonal component and horizontal component of Haar wavelet coefficients of mixed images. After screening SSPs, the signal features are sparser. The experiment results showed that the algorithm could estimate the mixing matrix faster and more accurately, and it could inspire to identify the latent variables by statistical histogram.

• 出版日期2013