Optimization Of Noisy Image Reconstruction Algorithm With Sparse Sampling

Sparse sampling projection data reconstruction algorithm is a technique to calculate reconstruction image by ray energy attenuation.It is mainly used in CT,X-ray machine,PET,Transmission electron microscope and other equipment.Taking CT as an example,the traditional CT reconstruction algorithms need complete projection data to get the ideal reconstruction results.However,it takes plenty of time to collect projection data of all angles and increase the amount of radiation,which is even not achievable in some situations.In the sparse sampling of projection data,in order to meet the limitation of radiation quantity,it is necessary to collect the projection data at the sparse angle.While in other situations,affected by scan environment and scan object,the obtained projection data are angle-limited.Both angle-sparse projection data and angle-limited projection data has brought great challenges to high quality reconstruction.Therefore,this paper discusses and optimizes the sparse sampling projection data reconstruction algorithm.The main contents are as follows:(1)For large data scale of angle-sparsed projection data and projection matrix,which led to large computation in single iteration,this paper proposes fast fusion reconstruction algorithm for multiscale projection data with sparse sampling.This method divide the iterative process into several steps,and obtains multi-scale projection data and projection matrix via the image pyramid,and for each scale,use different methods.Finally the merged results of each scale will be used as initial value to iterative reconstruction.Experimental results show that the fast fusion reconstruction algorithm for multiscale projection data under sparse sampling has a higher speed than traditional iterative algorithms,which also guarantees the reconstruction quality.(2)For angle-limited projection data reconstruction,considering the objective function is ill-posed,it is difficult to obtain the global optimal solution by conventional methods.So the reconstruction algorithm based on stochastic optimization is proposed.Through combining the stochastic optimization algorithm with traditional regularization(based reconstruction method),the reconstruction problem with angle-limited projection data is available.Meanwhile,the original stochastic optimization method is optimized to make it more adaptable to the reconstruction processes.Experimental results show that the reconstruction quality of this algorithm is better than those of other reconstruction algorithms.