SAR Images Compressed Sensing Based on Recovery Algorithms

We present in this paper a synthetic aperture radar compression proposal based on Compressed Sensing (CS). The major problem of CS when applying it on real data, is to make the processed data sparse. This process might be impossible without having predefined information about it to generate an adequate transform basis before the acquisition, which makes the CS unusable on scenes that present many scatters. The presented proposal is an alternative for finding a basis for every signal and it assures a sparse representation and a good reconstruction. We propose to use a recovery algorithm in the compression and the recovery steps of CS using the FFT basis as a representation matrix. We use several combinations of different measurement matrices and recovery algorithms, to see which one gives a fast processing and a good reconstruction. 30% of data are enough to recover the image well. This process is applied on Flevoland image and evaluated using correlation coefficient and processing time.