非局部稀疏表示正则化的磁共振图像重建

传统的基于稀疏性先验和全变分正则项约束的图像重建算法不能有效重建图像中的各种结构。为了提高重建质量,在采用传统重建算法中基于稀疏性的先验约束的同时,将重建图像的稀疏系数应逼近原始图像稀疏系数这一先验约束引入图像重建模型。通过图像子块之间的非局部相似性估计原始图像,得到非局部稀疏表示正则化的磁共振图像重建模型,并利用快速混合分裂算法求解模型。实验结果表明,算法能够对磁共振图像进行较好的重建。

MRI Reconstruction Based on Non-local Sparse Representation Regularization

The conventional algorithms based on effectively reconstruct various structures in images. use of the conventional sparsity prior as well as the coefficients and the sparse coefficients of original sparsity prior and total variation regularization can not To improve the performance of reconstruction, we make prior of the similarities between the reconstructed sparse image to regulate the model. Then the model is formed based on non-local similarity between patches and sparsity representation. Fast composite splitting algorithm is used to solve this model. According to the results of experiments, the proposed algorithm can achieve a better reconstruction performance for magnetic resonance imaging（MRI） images.