Research On Image Reconstruction Algorithm Based On Compressed Sensing

Traditional data compression technology is based on Nyquist sampling theorem,reaching the goal of compression by decreasing redundancy according to the speciality of information.As a newly-developing information processing technique,the CS technique breaks the constraint of traditional Nyquist sampling theorem thoroughly,it transforms signal acquisition into information acquision by taking advantage of signal's sparsity.The novelty of compressed sensing is that it samples dates and compresses dates simultaneously.CS technique chiefly consists of three respects,that is,the sparsity transformation of the signal,the design of the measurement matrix and the reconstruction algorithms.Because of the speciality of compressed sensing,applying this technique to some actual practices such as image fusion also obtains a good effect.This thesis mainly research the reconstruction algorithms and its application in image reconstruction and image fusion.The main contents of this thesis are as follows:1.Greedy algorithms can reconstruct signal precisely on the condition that the sparsity of the signal is known beforehand,in this context,we put forward a optimized greedy algorithm.This algorithm bases on GOMP algorithm,we use the part Hadamard measurement matrix instead of Gaussian measurement matrix and estimate sparsity by matching tests,meanwhile,we also use the threshold regularization theory so as to reconstruct signal precisely.By contrasting the experimental results,we can find that compared with other algorithms,this algorithm has a more excellent recovery performance and a higher reconstruction probability under the same conditions.2.There are many kinds of image fusion methods based on CS theory,the typical one among them is a method based on orthogonal matching pursuit algorithm.In this thesis,We do some experiments based on the typical method and the new method respectively.Experimental results show that the new reconstruction algorithm has a higher reconstruction precision than the typical method under the same conditions,furthermore,this optimized algorithm can promote the image gain a better fusion result.