Research On Data-Driven Sparse Angle CT Reconstruction Algorithm

Computed Tomography（CT）imaging technology is recognized as one of the greatest scientific and technological achievements nearly a hundred years.This technology not only has a revolutionary impact on medical diagnosis,but also has been successfully applied to non-destructive testing of industrial,geological exploration,resource exploration,etc.Although many CT systems have been extended in various industries,their theoretical core is mainly image reconstruction.However,with the continuous development of practical applications,various industries have put forward many new requirements for CT reconstruction.For example,the double-deep well mode is often used in site exploration.One deep well discharges the detector and one deep well discharges the transmitter.It is absolutely impossible to get the theoretical complete angle numbers.Degree collection.In addition,the acquisition of medical images in hospitals often relies on X-rays and other rays that are harmful to the human body.Excessive radiation can cause various diseases and seriously endanger the health of patients,At the same time,the time-consuming acquisition of full-angle data and the space of the device restrict the using of CT reconstruction technology.Sparse angle scanning is an important technical means to shorten the acquisition time and reduce the amount of radiation received by the human body.However,the reduction of the projection angle numbers will lead to serious degradation of the reconstruction results.Therefore,it is necessary to study how to optimize the image reconstruction results under the sparse angle data acquisition.research on how to get great reconstruction by using the data collected under the number of sparse angles has an important meaning in promoting the using of sparse angle CT technology.This paper will introduce the current popular algorithms for finding the optimal filter and conduct research on this basis,and research and explore new algorithms for optimizing image reconstruction.The main research work is as follows:（1）The optimal filter solution algorithm based on Nonlinear Neighborhood Filters（NNFs）is studied.Traditional filtering often does not make good use of pixel similarity to determine the filtering weight,and the filtering result will lead to the blurring of detailed information.This paper will study this problem,discuss several common nonlinear neighborhood filters（NNFs）methods,and add them to the image reconstruction algorithm to find the optimal filter by deriving appropriate mathematical formulas in the form of regular terms.The new algorithm makes full use of the advantages of nonlinear domain filtering（NNFs）,so that the smoothing operation that depends on the similarity of the neighboring pixels to determine the characteristics of the filter matrix weight is added in the process of solving the optimal filter.The optimal filter for filtering back-projection will obtain better reconstruction results,and at the same time,it has the advantage of finding the optimal filter in terms of time-consuming.Finally,the performance of the proposed algorithm is verified by experiments,and the reconstruction results of projections from different angles and the increase in the algorithm time-consuming compared with the original algorithm are analyzed.The subjective visual effects and objective evaluation indicators show that the new algorithm can obtain better denoising.（2）For the non-local mean filter（NLM）in NNFs,a rotation-based non-local mean filter optimization scheme is proposed.When the similarity corresponding to all patches is concerned,the different angular positions of similar structural information are not considered.The simple NLM only compares all the uncertain similarities between the center patch and the search box.The algorithm optimizes by rotating the center patch to compare the similarity with the patch in the search box multiple times,by comparing the similarity between the patches under each group of rotation angles,the maximum similarity is selected as the calculation criterion,which will improve the Performance of NLM.The performance of the algorithm optimization is verified by appropriate algorithm programming,and the reconstruction results of different projections and the time-consuming increase of the algorithm are analyzed.The subjective visual results and objective evaluation indicators show that the new algorithm will obtain better denoising effect and better preserve the structural information of the original image.