The CT Image Reconstruction Algorithm Based On Dictionary Learning Research

In the field of image reconstruction for CT, there are already many studies onreconstructing methods, which has made an obvious achievements. Compressed sensingis an emerging theory about information acquisition and processing, which could make usacquire high-quality reconstructive images by virtue of less data. In recent years,compressed sensing has a very prominent effect on the improvement of medical imagereconstruction algorithm. As one of image sparse representation’s optimal algorithms inthe field of compressed sensing, dictionary training algorithm updates dictionary byvirtue of sparse decomposition and training, which can more preferably adapt ormore effectively capture image features than other fixed sparse modes.In the first place, this paper expounds the principles of CT imaging, iterative method,analytic method and two methods of objectively evaluating the quality of imagereconstruction, compares several iterative algorithms, and shows effects of differentreconstruction algorithm by virtue of simulation experiments.In the second place, on the basis of least-absolute criteria, this paper converts imagereconstruction problem into optimization problem, improves simplistic iterative method,and verifies effectiveness of the iterative method by virtue of experiments.In the third place, this paper expounds some basic theories of dictionary training,systematically describes dictionary training being applied to image reconstruction,combines K-SVD dictionary training algorithm with ART algorithm, based on the imageblock structure dictionary, and conducts sparse decomposition by virtue of the OMPalgorithm. Experimental results show that the algorithm can reconstruct the image byvirtue of less projection data, select some appropriate key parameters of iterativealgorithm itself, and finally obtain high-quality reconstructive images.