关键词: 压缩感知, 快速重构, 最小二乘优化, 整体相关参数, 分块重构

Abstract:

The new sampling method of compressed sensing (CS) can sample signals at a very low rate that is well below the predetermined sampling rate set by the Shannon-Nyquist sampling theorem. But the reconstruction time of CS algorithms is longer and it grows rapidly with the size of signals. To overcome the abovementioned problems, we propose an algorithm, called fast block whole reconstruction for image (FBWRFI), which can quickly reconstruct image signals. The FBWRFI reconstructs signals based on the least squares method and chooses the most relevant atoms by using the overall-correlated parameters. We also introduce the theory of block reconstruction and redesign the size of blocks and the measurement matrixes. Theoretically, the FBWRFI can reduce the computation complexity and computing scale by a large margin. Experimental results show that the FBWRFI algorithm can significantly reduce reconstruction time and the growth rate of reconstruction time changing with the size of signals, which demonstrates the proposal's effectiveness.

Key words: compressed sensing, fast reconstructed, the optimized method of least-square, overall-correlated parameter, block reconstruction