Ke Jun, Lam Edmund Y
School of Optoelectronic, Beijing Institute of Technology, Beijing, China.
Opt Express. 2012 Sep 24;20(20):22102-17. doi: 10.1364/OE.20.022102.
A block-based compressive imaging (BCI) system using sequential architecture is presented in this paper. Feature measurements are collected using the principal component analysis (PCA) projection. The linear Wiener operator and a nonlinear method based on the Field-of-Expert (FoE) prior model are used for object reconstruction. Experimental results are given to demonstrate the superior reconstruction performance of the FoE-based method over the Wiener operator. In addition, the effects of system parameters, such as the object block size, the number of features per block, and the noise level to the BCI reconstruction performance are discussed with different kinds of objects. Then an optimal block size is defined and studied for BCI.
本文提出了一种采用顺序架构的基于块的压缩成像（BCI）系统。使用主成分分析（PCA）投影来收集特征测量值。线性维纳算子和基于专家场（FoE）先验模型的非线性方法用于目标重建。给出了实验结果，以证明基于FoE的方法相对于维纳算子具有卓越的重建性能。此外，针对不同类型的目标，讨论了系统参数（如目标块大小、每个块的特征数量以及噪声水平）对BCI重建性能的影响。然后定义并研究了BCI的最优块大小。
