Joint dictionary training optimization method for MR image super-resolution

The invention discloses a joint dictionary training optimization method for MR image super-resolution, and belongs to the field of MR image reconstruction. The method comprises the following steps: 1, inputting a training image block pair dictionary size K which represents a training sample of a space X and is a training sample of a space Y; 2, initializing the sample, wherein n is equal to 0, and t is equal to 1; 3, performing circulation and executing the following optimization formula for i = 1, 2,..., N; 4, updating the merged column vector, and ending the circulation; 5, updating according to the following formula; 6, outputting the dictionary pair. According to the method, the reconstruction error is effectively reduced, and the image reconstruction precision is improved. 本发明公开了一种面向MR图像超分辨的联合字典训练优化方法，属于MR图像重建领域。步骤一：输入训练图像块对字典大小K，其中，表示空间X的训练样本，是空间Y的训练样本；步骤二：初始化和n＝0，t＝1；步骤三：循环，对于i＝1,2,...,N执行如下优化公式；且步骤四：更新归一下的列向量，结束循环；步骤五：根据下面公式更新步骤六：输出字典对和本方法有效减小了重构的误差，提升了图像的重构精度。