Research On Structured Measurement Matrix And Coding Algorithm In Compressed Sensing

In compressed sensing, the image is sampled by measurement matrix at a single sampling rate. Measurement matrix plays important roles in signal acquisition and reconstruction. The traditional measurement matrix could achieve good performance on image reconstruction, but usually occupies numerous system resources because of a large number of sampling. Compared with random measurement matrices, the deterministic measurement matrices possess their own constraints, which lead to worse performance. In order to solve these problems, two kinds of structured random matrices and a multi-layered block adaptive coding algorithm are proposed.Based on the generalized rotation matrix, two kinds of structured random matrices are proposed, which are the generalized binary rotation matrix and the pseudo-random generalized binary rotation matrix. These two series of new matrices perform better than the traditional measurement matrices. The amount of time required by the traditional and new approaches is about the same. What's more, they could obtain more accurate reconstructions at low sampling rates.On the basis of the adaptive BCS algorithm(ABCS) using the block OSTM, a multi-layered block adaptive coding algorithm(MLBA) and a multi-layered block adaptive compressed sensing codec method(MLBACS) are proposed. According to its regional structure, the image is divided into blocks of different layers and different sizes, and each block is allocated a different sampling rate. The sampling number of MLBACS is less than traditional compressed sensing at a single sampling rate, with the same reconstruction performance. Compared with ABCS, MLBACS breaks through the limitation of the specific measurement matrix, and possesses the advantage of processing the images with a large smoothed area.The research on these two areas could be widely applied to different kinds of fields, and deserves further study in the coming years.