Research On Blocked High Resolution Compressed Imaging

In this paper,Compressed Sensing(CS)is used to combine the signal sampling and compression process to obtain high resolution images that exceed the resolution of the hardware camera.Compressed Sensing takes advantage of the sparse characteristics of the target and records the image in a much smaller amount than the Nyquist sampling theorem.The traditional compressed sensing uses a single pixel camera to collect data,but it will meet the problems of long acquisition time and a large amount of data in the recovery process,which leads to high memory occupancy and long recovery time.In order to solve the above problems,Block Compressed Sensing(BCI)is used in this thesis.An array camera is used instead of a single pixel camera to reduce the number of sampling times.At the same time,a small block is used to restore the target to reduce the amount of data,thus reducing the memory footprint and shortening the recovery time.However,with the increasing of resolution,the speed of restoration algorithm is the main factor limiting the speed of BCI.In order to speed up the process,in addition to optimizing the structure of the algorithm,this paper uses GPU parallel computing to accelerate the algorithm.Firstly,this paper studies the principle and method of block compression imaging,including sparse representation of signal,construction of measurement matrix and recovery algorithm.The concrete realization steps of the four restoration algorithms are studied and the recovery effects are observed through simulation experiments.At the same time,the factors that affect the final recovery effect are analyzed in detail.Secondly,the compression imaging experiment platform is designed and built in this paper.The data acquisition process of the block compression imaging is completed by loading the matrix modulated image with DMD and receiving the modulated image by the CCD camera.Then four restoration algorithms are used to restore the image.Finally,in view of the slow speed of the block compression imaging recovery algorithm,this paper uses GPU parallel computing technology to accelerate the recovery algorithm on two levels,the first one is the parallel acceleration of the matrix operation and the second one is the parallel acceleration between each block.The accelerated algorithm significantly reduces the time used to compute high resolution images.