摘要
针对压缩感知子空间追踪SP(subspace pursuit)算法必须以信号稀疏度为先验知识,而现实中图像稀疏度未知这一问题,提出改进SP算法MSP(modified subspace pursuit)。首先对信号的稀疏度进行自适应估计,其次在迭代过程中,通过给定的步长因子对稀疏度进行更新,使之逐渐逼近正确子空间,当重构误差小于阈值时,停止迭代,实现稀疏信号的重构。重构图像表明:MSP算法在运算时间和重构精度上均优于其他同类算法,实现了图像的快速精确重构。
Compressive sensing subspace pursuit (SP) has to take the sparsity of signals as its priori knowledge, while the image sparsity actuality cannot be firstly got. Aiming at this problem, the modified-SP (MSP) algorithm is proposed in this paper. First, it adaptively estimates the sparsity of the signal; then in iteration process, it updates the sparsity by the given step factor to enable gradually approaching to the correct subspace, and stops iterating when the reconstruction error is less than the threshold, thus achieves the reconstruction of sparse signal. Reconstructed images show that the MSP algorithm outperforms other similar algorithms in computation time and reconstruction accuracy, and realises fast and accurate image reconstruction.
出处
《计算机应用与软件》
CSCD
北大核心
2013年第7期200-203,216,共5页
Computer Applications and Software
基金
陕西省科技厅科技攻关项目(2012K06-16
2011K09-36)
陕西省教育厅科学研究计划项目(12JK0528)
关键词
压缩感知
子空间追踪
重构
图像处理
采样
稀疏度
Compressive sensing Subspace pursuit Reconstruction Image processing Sampling Sparsity