压缩感知OMP算法与IRLS算法在计算鬼成像中的对比分析

Contrast Analysis of Compressive Sensing OMP and IRLS Algorithm in the Computational Ghost Imaging

将压缩感知技术和鬼成像系统相结合,能够大幅度地降低成像所需的测量次数,并能有效地提高重构图像的峰值信噪比。本文将离散余弦变换(DCT)矩阵作为图像稀疏化矩阵,采用正交匹配追踪算法(OMP)和迭代加权最小二乘算法(IRLS)两种压缩感知算法作为压缩感知鬼成像系统图像重构的算法。通过对两种算法在改变稀疏度和测量次数时重构结果的峰值信噪比变化的比较,探究了这两个变量对峰值信噪比的影响。发现IRLS算法重构精度更高,图像质量更好,而OMP算法迭代速度比IRLS更快,重构图像所需的时间较少。

The number of measurements can be reduced greatly by combining the compressive sensing technology with ghost imaging system. It can also effectively increase the peak signal to noise ratio of the reconstructed image. In this paper, the discrete cosine transform matrix is used as the image sparse matrix. Two kinds of compressive sensing algorithm, the orthogonal matching pursuit algorithm and the iterative weighted least square algorithm, are used as the com- pressed sensing image reconstruction algorithm. We consider the influence of the number of measurements and the sparsity to peak signal to noise ratio by comparing the variety of peak signal to noise ratio with various values of sparsity and measurements. It is found that the reconstruction accuracy of IRLS algorithm is higher, the image quality is better, and the OMP algorithm is faster than IRLS, and the required time for the reconstruction is less.