Study On Reconstrution Of Compressed Sensing Based On Particle Swarm Optimization And Levenberg-marquardt

The traditional sampling of signal must obey the Shannon theorem, so the frequencyof sampling signal must be at least twice of original signal to avoid losing the informationof signal, but further improving the Nyquist frequency will increase the complexity fordata capturing. Compressed sensing theory presents a new method to capture andcompress the data, which applies the sparsity prior of the signals or images and canaccurately reconstruct original signals or images from a small quantity of measurements.Aimed at the disadvantage of the slow recovered speed and the bad recovered quality, weapply the compressed sensing to the signal and image reconstruction based on the existingmethods, and mainly research on the following aspects in this paper.Firstly, the current and mainstream compressed sensing and reconstructionalgorithms have been induced and summarized. Making use of the sparsity of image, westudy a reconstruction method based on total variation, and then perform the experimentson total variation andl 2function and compare their performance.Secondly，because of premature convergence of particle swarm optimization anddependence on the initial conditions in the use of Levenberg-Marquardt algorithm, wepropose a new hybrid intelligent algorithm of using alternately particle swarmoptimization algorithm and Levenberg-Marquardt algorithm, and we compare hybridintelligent algorithm with particle swarm optimization algorithm and Levenberg-Marquardt algorithm. The experiment results show the validity of the hybrid intelligentalgorithm.Thirdly，matching pursuit algorithm is analyzed for signal decomposition in thispaper, for the huge computational problem of the algorithm, we make use of a new hybridintelligent algorithm using alternately particle swarm optimization algorithm andLevenberg-Marquardt algorithm to realize the signals decomposition, then we comparehybrid intelligent algorithm with matching pursuit algorithm. The experiment shows thatthe hybrid algorithm considerably reduces the computation time.At last, orthogonal matching pursuit and particle swarm algorithm are analyzed in this paper, for the bad recovered quality of the algorithms, so we discuss the algorithm thatone can reconstruct the image using particle swarm optimization algorithm andLevenberg-Marquardt algorithm, and compare the hybrid algorithm with orthogonalmatching pursuit and particle swarm algorithm.The results of experiments show that ouralgorithm effectively improves the quality of the recovered image.