Foreground and background separation of sparse infrared image based on combining sensing matrix

A method based on compressed sampling and matrix decomposition to separate foreground and background of infrared sparse images was proposed in this article. A new combinational sensing matrix was used to obtain measurements of original infrared images. The thumbnail and the compressed sampled values are obtained by the combinational sensing matrix at the same time. Rank estimation based on image information entropy and sparse recovery was used to reconstruct foreground and background in the compressed domain. Experiments was carried out on real infrared images. Compared with MaxMedian, TopHat and CLSDM, the signal-to-noise ratio gain and background suppression factor have been significantly improved by using our method.