基于改进StOMP算法图像压缩感知重构*

在图像压缩感知重建中，一些算法能够取得好的重构效果，但耗时较长；一些算法耗时较短，但又不能取得精确解。针对重构效果和耗时不能兼得的问题，本文基于小波域稀疏，选用常规观测矩阵进行观测采样，通过对观测结果预定义滤波、选取信号硬阈值，引入共轭梯度下降算法，对分段正交匹配追踪（StOMP）重建算法进行改进。提出重建图像的边缘相似度概念，并对不同压缩比下的观测信号重建进行实验仿真。结果表明，相对于改进前StOMP算法，改进后StOMP算法在迭代收敛时间较短的情况下，重构效果提升。在主观评价上，重建图像噪声点明显减少；客观评价上，PSNR值提高，达到预期效果。

Image reconstruction for compressed sensing based on improvement of StOMP algorithm

In the Image reconstruction for compressed sensing, some algorithms can achieve good reconstruction effect, but take longer time.Some algorithms take shorter time, but they can''t get the exact solution.The conventional observation matrix was used to take compression sampling based on sparse Wavelet Domain in this paper.To improve Stagewise Orthogonal Matching Pursuit(StOMP),The conjugate gradient descent algorithm was introduced through the predefined filtering of the observation results and the hard threshold of the signal.The concept of edge similarity of reconstructed image was proposed and the reconstruction of observation signal under different compression ratio is simulated.The results show that the reconstruction effect is improved compared with the improved StOMP algorithm whose iterative convergence time is still short.In the subjective evaluation, the noise points of the reconstructed images are obviously decreased;Objectively, the Power Signal-to-Noise Ratio(PSNR) value is improved and the expected effect is reached.