Compressed Sensing Algorithm Based On Adaptive Restart

With the emergence of compressed sensing (CS), a significant amount of efforts have been put forward to develop both the related scientific research and industry application. CS tries to use the fewest measurements to obtain the sparse representation of the original signal in the end. Consequently, on the data collection, storage and transmission, CS provides much convenience to people’s lives.This article will start from background introduction of compressed sensing, and then introduce its current research status in various fields. There are a lot of reconstruction algorithm of compressed sensing and its improved method. In this paper, some algorithms are arranged according to their basic principles of each method and each typical algorithm is introduced. At last, computing complexity of these methods are compared.As a kind of important reconstruction methods of compression sensing, NESTA algorithm’s complexity is far lower than the other two order optimization algorithm. At the same time, in the first order algorithm, its accuracy is relatively high. In addition, NESTA as a tool, is widely used in other optimization problems.The thoughts of restart algorithm has a lot of advantages in solving the convex optimization problem with a smooth function, especially for compression sensing as well as this type of sparse optimization problem.NESAT is insufficient in the process of iteration convergence. Inspired by restart algorithm, R-NESTA is proposed in this paper based on the relative variation adaptive restart. Through a set of simulated data and real data simulation experiment, the modified R-NESTA algorithm effectively makes up for NESTA iterative shortcomings of slow convergence.FAN-Large database is a large comprehensive database of images and the corresponding captions. Its main feature is the ability to find the face in the images corresponding the names in the captions. This article will apply R-NESTA into face detection applications in this database. Compared with face detection routines in Opencv, part of the FAN-Large database’s image retrieval effect are good.