Compressed Sensing System Based On Aerial Imagery

In recent years, the compressed sensing theory breaks the limitations of traditional Nyquist sampling theorem for reconstructing sparse or compress signals by using a small amount of signal values. Moreover, it successfully overcomes the resistance such as the collection of huge amount of sampled data, the long sampling time, and the serious resource waste for data storage and transmission. Compared to natural images, aerial images are often characterized with strong randomness, weak local correlation, and multiple small targets. The existing block compressed sensing algorithms failed to fully explore these characteristics. Based on these characteristics of aerial imagery, new conclusions and algorithms will be present in this thesis.After carefully reviewing the application prospect, theoretical basis and research progress of image compressed sensing, this thesis analyzes and studies the block compression sensing system. Firstly, this thesis describes several common reconstruction algorithms, and compares their advantages and disadvantages in the performance simulations. On the foundation of BCS-SPL, this thesis proposes a block compressed sensing reconstruction algorithm based on the adaptive-thresholding iteration: block compressed sensing with adaptive-thresholding projected Landweber, which leverages a piecewise-linear thresholding model for bivariate shrinkage denoising and global projected-Landweber iterative reconstruction. Then, a system of block measurement & global reconstruction is proposed, where the decoder reconstructs the whole image by introducing the reordering operator, and thus it can avoid the blocking artifacts from block-by-block reconstruction. Finally, based on the block measurement & global reconstruction system, the block-by-block adaptive measurement is proposed to guarantee more efficient measurement for further improving the performance of the whole system.