摘要
稀疏度自适应匹配追踪(SAMP)算法是压缩感知(CS)中一种主流的图像重构算法。随着迭代次数的增加,SAMP算法的原子候选集将成倍增加,会导致系统空间的浪费和重构时间的增长。为此,提出了一种模糊裁剪阈值稀疏度自适应匹配追踪(FPTSAMP)算法。由于离散小波变换(DWT)在CS稀疏处理过程中破坏了低频逼近系数间的相关性,对信号的重构质量产生了一定的负面影响,因而采用小波高频子带变换(HFSBWT)来替代DWT,实现对信号的稀疏表示。仿真实验结果表明,相比于同一重构算法,采用HFSBWT方法得到的峰值信噪比更好;与SAMP算法相比,与HFSBWT相结合的FPTSAMP算法的重构效果有了明显提高,重构时间也减少了一半。
Sparsity Adaptive Matching Pursuit （SAMP） is a mainstream image reconstruction algorithm in Compressed Sensing （CS）. However,with the increase of iterative times,it has multiplied atoms candidate set that lead to wasting storage capacities and lengthening reconstruction time. A method called Fuzzy Pruning Threshold Sparsity Adaptive Matching Pursuit （FPTSAMP） is proposed. The Dis- crete Wavelet Transform （DWT） destroys the correlation among low-frequency approximation coefficients in CS sparsity processing, which results in bad reconstruction quality, so a High Frequency Sub-Band Wavelet Transform （HFSBWT） is adopted instead of DWT to realize the sparse representation of signal. Simulation results show that compared with the same reconstruction algorithms the HFSBWT has achieved a better Peak Signal To Noise Ratio （PSNR） of images and that compared with SAMP algorithm the FPTSAMP combined with HFSBWT has lifted the reconstruction performance of images significantly with its reconstruction time cutting in half.
出处
《计算机技术与发展》
2017年第9期35-39,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(61501251,61071167,61373137)
江苏省普通高校研究生科研创新计划资助项目(KYZZ15_0236)
南京邮电大学引进人才科研启动基金资助项目(NY214191)
关键词
压缩感知
重构算法
高频子带小波变换
模糊裁剪阈值SAMP算法
compressed sensing
reconstruction algorithm
high frequency sub-band wavelet transform
fuzzy pruning threshold sparsity adaptive matching pursuit
