Kernel ridge regression-based TV regularization for motion correction of dynamic MRI

•The KRR-based TV model is studied to dynamic MR image reconstruction. Different from the classic TV-based model, we exploit KRR technique to establish the relationship between the current and the reference frame, which can efficiently correct the corrupted reference frame.•To preserve the spatial details in reconstruction process, the gradient-domain-guided filtering is incorporated in the KRR-based TV model, which can preserve image edges and deblocking artifacts in the reconstructed image.•The the split Bregman iteration (SBI) method is utilized and extended to solve the proposed model. Simulations are performed to validate the effectiveness of the proposed method, in comparison with some of the aforementioned techniques, in both the quantitative evaluation and the visual comparison.