基于改进StOMP算法图像压缩感知重构

Image reconstruction for compressed sensing based on improvement of StOMP algorithm

在图像压缩感知重建中,针对重构效果和耗时不能兼得的问题进行深入研究。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(St OMP)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St OMP算法,改进算法在迭代收敛时间较短的情况下,重构效果有所提升。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,达到了预期效果。该算法为高效解决压缩感知优化重建问题提供了参考。

To solve the problem that good reconstruction effect and short time consuming could not achieve both at the same time in the image reconstruction for compressed sensing,this paper took an in-depth research. The processing used the conventional observation matrix to take compression sampling based on sparse wavelet domain. To improve stagewise orthogonal matching pursuit( St OMP),this paper introduced the conjugate gradient descent algorithm after finishing the predefined filtering of the observation results and the hard threshold of the signal. It proposed the concept of edge similarity of reconstructed image and simulated the reconstruction of observation signal under different compression ratio. The results show the improved algorithms whose iterative convergence time is still short promotes reconstruction effect. In the subjective evaluation,the algorithm obviously reduces the noise points of the reconstructed images. Objectively,the algorithm improves the power signal-tonoise ratio( PSNR) value and reaches the expected effect. This algorithm provides a reference for the efficient solution of the compressed sensing optimization problem.