Improved morphological component analysis for interference hyperspectral image decomposition

Due to the special imaging principle, lots of vertical interference stripes exist in the frames of the IHI (interference hyperspectral image) data, which will affect the result of compressed sensing theory or other traditional compression algorithms used on IHI data. In this paper, MCA (morphological component analysis) algorithm is adopted to separate the interference stripes layers and the background layers, and an IMCA (improved MCA) algorithm is proposed according to the characteristics of the IHI data, dictionary learned from the LSMIS (Large Spatially Modulated Interference Spectral Image) data is used to sparsely represent the stripes layers instead of traditional basis, and the condition of iteration convergence is improved. The experimental results prove that the proposed IMCA algorithm can get better results than the traditional MCA, and also can meet the convergence conditions much faster than the traditional MCA.