Research On Image Reconstruction Algorithm Of Compressed Sensing

With the further expansion of intelligent sensor network and the countless sensor products coming into practical service in the 21 st century, high speed image and information processing is facing enormous challenges, and the data collected by sensor devices will be unprecedented. In the 20 th century, the vast majority of collecting signal of sensor devices, mainly based on Shannon- Nyquist sampling theorem for theoretical guidance. By Shannon- Nyquist sampling theorem specifies information sampling rate of the minimum value, so it becomes a necessary condition for accurately reconstruct the original data signal. In recent years, broke through the Shannon- Nyquist sampling theorem of compressed sensing technology is put forward, it is no longer restricted by the lowest sampling frequency information, marked a milestone in the history of data acquisition. Based on compression perception theory research in the theory of CS sample rate, how to decreases and to ascend the rate of refactoring, for the development in the field of signal processing of new methods, raised the sparse degree of signal and the signal of the reduction rate. After the study of classical reconstruction algorithm analysis, this paper proposes two algorithms: first, the improved orthogonal matching pursuit algorithm; Second, an image reconstruction algorithm based on Stepwise Regularization Subspace Pursuit algorithm in compressed sensing. In this paper, the content of the main research work is as follows:An improved orthogonal matching pursuit algorithm is put forward. After a series analysis most of the existing orthogonal matching pursuit algorithm, the common deficiency is that, in the signal reconstruction phase, on the premise of need to input the original signal sparse, and reconstruct the insufficient problems such as low precision, accuracy, this paper puts forward an improved orthogonal matching pursuit algorithm, Its main feature is to signal sparse degree and support set with adaptive features a collaboration of reconstructing signal method. Specific steps are as follows: sparse degree by prediction information, collaborative support updating and extension set, reduce the error estimation and error correction of support set. Improved orthogonal matching pursuit algorithm not only is more superior than the OMP algorithm on image restoration, in terms of reconstruction error can also reduce with the increase of measurement signal dimension, and is more obvious than the OMP algorithm on the downward trend.Putting forward an image reconstruction algorithm based on Stepwise Regularization Subspace Pursuit algorithm in compressed sensing. Its main idea is using the method of threshold processing as an out support concentration of atoms before iteration algorithm, and atoms of regularization processing candidate set. Experiments show that when the sampling rate is 0.5, the effect of image reconstruction PSNR and reconstruction time of SRSP is better than other compared with the algorithm(MP, OMP, StOMP, ROMP, CoSaMP and SP).