Motion correction improves the sensitivity to slow blood flow and the resolution in super-resolution imaging pipelines. The motion from breathing and heartbeats brings images out of alignment, but it can be estimated and corrected by resampling the images using 2D interpolation. However, MATLAB's high-quality interpolation method, "spline", lacks a GPU implementation for 2D images, leading to long processing times. This conference paper, with the code included, shows how to implement 2D spline interpolation, enabling real-time motion correction at a rate of 5000 kHz using a standard GPU.
https://orbit.dtu.dk/en/publications/fast-spline-interpolation-using-gpu-acceleration
