| Computed tomography(CT)is an advanced medical image assisted diagnosis technology with high temporal and spatial resolution,which has been widely used in the health examination of hospitals at home and abroad.However,related studies have shown that excessive radiation can increase the risk of cancer and other hereditary diseases.Sparse angle sampling is considered as an effective way to reduce the radiation dose,which can not only reduce the radiation damage by reducing the number of X-rays passing through the human body,but also prolong the service life of the equipment.However,the reduction of sampling angle means that the projection data does not meet the requirements of completeness.Sparse Angle imaging will inevitably lead to the degradation of reconstructed image quality.Therefore,how to reconstruct CT images with better quality under sparse Angle sampling plays an important role in promoting the wider application of sparse angle CT imaging technology.In this paper,aiming at the problem of image quality degradation of CT reconstruction under sparse angle sampling,two CT iterative reconstruction algorithms based on sparse representation are studied.Some preliminary progress and results have been made.The main work of this paper is as follows:(1)A sparse angle CT reconstruction method based on group sparse representation is proposed.The traditional dictionary learning method is processed block by block,ignoring the self-similarity of the image itself,which may make the sparse representation of similar image patches completely different.The group-based sparse representation no longer takes a single image patch as the minimum unit of sparse representation,but uses Euclidean distance as the similarity measure,divides similar image patches into a group,and takes the group as the basic unit of sparse representation.This method fully considers the local sparsity and non-local selfsimilarity of the image.In this paper,the sparse representation based on groups is introduced into the punishment least square iterative reconstruction framework as a regular term to construct the objective function.Multiple variables in the objective function are solved by alternate minimization,and the final reconstruction results are obtained.The qualitative and quantitative evaluation results of several groups of experimental results prove the advantages of the proposed algorithm,which can not only remove the bar artifacts,but also effectively protect the image details.(2)A sparse angle CT reconstruction algorithm based on deep convolution sparse representation is proposed.Convolution sparse representation is used in many image processing fields because of its translation invariance,good interpretability and better detail and texture representation.In this paper,the convolutional sparse representation is taken as the regularization constraint of iterative reconstruction.Convolutional neural network(CNN)is used to realize convolutional sparse coding and convolutional dictionary learning.In detail,the method is divided into two sub-problems,which correspond to two solving processes:reconstruction process and deep convolution sparse representation regularization constraint process.Firstly,the intermediate reconstruction image is obtained by using synchronous algebraic reconstruction technique(SART),and then the convolution sparse representation model based on intermediate reconstruction image is established by using deep convolution sparse representation algorithm.Specifically,the deep convolution sparse representation combines the numerical solution with the training based on deep learning,alternately learns the convolution dictionary and convolution sparse feature mapping,and uses CNN to parameterize the generalized a priori in the convolution sparse representation model.The final reconstructed image is obtained by alternating SART and deep convolution sparse representation.The experimental results show that the image reconstructed by the proposed algorithm has good tissue texture and obvious artifact suppression effect,and is better than the contrast method in quantitative and qualitative analysis. |
