The Wavelet-Domain Projection Pursuit Learning Network (WDPPLN) is proposedfor restoring degraded image. The new network combines the advantages of both projectionpursuit and wavelet shrinkage. Restoring image is very...The Wavelet-Domain Projection Pursuit Learning Network (WDPPLN) is proposedfor restoring degraded image. The new network combines the advantages of both projectionpursuit and wavelet shrinkage. Restoring image is very difficult when little is known about apriori knowledge for multisource degraded factors. WDPPLN successfully resolves this problemby separately processing wavelet coefficients and scale coefficients. Parameters in WDPPLN,which are used to simulate degraded factors, are estimated via WDPPLN training, using scalecoefficients. Also, WDPPLN uses soft-threshold of wavelet shrinkage technique to suppress noisein three high frequency subbands. The new method is compared with the traditional methodsand the Projection Pursuit Learning Network (PPLN) method. Experimental results demonstratethat it is an effective method for unsupervised restoring degraded image.展开更多
This paper aims to meet the requirements of reducing the scanning time of magnetic resonance imaging （MRI）, accelerating MRI and reconstructing a high quality image from less acquisition data as much as possible. MR...This paper aims to meet the requirements of reducing the scanning time of magnetic resonance imaging （MRI）, accelerating MRI and reconstructing a high quality image from less acquisition data as much as possible. MRI method based on compressed sensing （CS） with multiple regularizations （two regularizations including total variation （TV） norm and L1 norm or three regularizations consisting of total variation, L1 norm and wavelet tree structure） is proposed in this paper, which is implemented by applying split augmented lagrangian shrinkage algorithm （SALSA）. To solve magnetic resonance image reconstruction problems with linear combinations of total variation and L1 norm, we utilized composite spht denoising （CSD） to split the original complex problem into TV norm and L1 norm regularization subproblems which were simple and easy to be solved respectively in this paper. The reconstructed image was obtained from the weighted average of solutions from two subprohlems in an iterative framework. Because each of the splitted subproblems can be regarded as MRI model based on CS with single regularization, and for solving the kind of model, split augmented lagrange algorithm has advantage over existing fast algorithm such as fast iterative shrinkage thresholding（FIST） and two step iterative shrinkage thresholding （TWIST） in convergence speed. Therefore, we proposed to adopt SALSA to solve the subproblems. Moreover, in order to solve magnetic resonance image reconstruction problems with linear combinations of total variation, L1 norm and wavelet tree structure, we can split the original problem into three subproblems in the same manner, which can be processed by existing iteration scheme. A great deal of experimental results show that the proposed methods can effectively reconstruct the original image. Compared with existing algorithms such as TVCMRI, RecPF, CSA, FCSA and WaTMRI, the proposed methods have greatly improved the quality of the reconstructed images and have better visual effect.展开更多
文摘The Wavelet-Domain Projection Pursuit Learning Network (WDPPLN) is proposedfor restoring degraded image. The new network combines the advantages of both projectionpursuit and wavelet shrinkage. Restoring image is very difficult when little is known about apriori knowledge for multisource degraded factors. WDPPLN successfully resolves this problemby separately processing wavelet coefficients and scale coefficients. Parameters in WDPPLN,which are used to simulate degraded factors, are estimated via WDPPLN training, using scalecoefficients. Also, WDPPLN uses soft-threshold of wavelet shrinkage technique to suppress noisein three high frequency subbands. The new method is compared with the traditional methodsand the Projection Pursuit Learning Network (PPLN) method. Experimental results demonstratethat it is an effective method for unsupervised restoring degraded image.
基金Natural Science Foundation of Chinagrant number:81371635+3 种基金Research Fund for the Doctoral Program of Higher Education of Chinagrant number:20120131110062Shandong Province Science and Technology Development Plangrant number:2013GGX10104
文摘This paper aims to meet the requirements of reducing the scanning time of magnetic resonance imaging （MRI）, accelerating MRI and reconstructing a high quality image from less acquisition data as much as possible. MRI method based on compressed sensing （CS） with multiple regularizations （two regularizations including total variation （TV） norm and L1 norm or three regularizations consisting of total variation, L1 norm and wavelet tree structure） is proposed in this paper, which is implemented by applying split augmented lagrangian shrinkage algorithm （SALSA）. To solve magnetic resonance image reconstruction problems with linear combinations of total variation and L1 norm, we utilized composite spht denoising （CSD） to split the original complex problem into TV norm and L1 norm regularization subproblems which were simple and easy to be solved respectively in this paper. The reconstructed image was obtained from the weighted average of solutions from two subprohlems in an iterative framework. Because each of the splitted subproblems can be regarded as MRI model based on CS with single regularization, and for solving the kind of model, split augmented lagrange algorithm has advantage over existing fast algorithm such as fast iterative shrinkage thresholding（FIST） and two step iterative shrinkage thresholding （TWIST） in convergence speed. Therefore, we proposed to adopt SALSA to solve the subproblems. Moreover, in order to solve magnetic resonance image reconstruction problems with linear combinations of total variation, L1 norm and wavelet tree structure, we can split the original problem into three subproblems in the same manner, which can be processed by existing iteration scheme. A great deal of experimental results show that the proposed methods can effectively reconstruct the original image. Compared with existing algorithms such as TVCMRI, RecPF, CSA, FCSA and WaTMRI, the proposed methods have greatly improved the quality of the reconstructed images and have better visual effect.
