Super-Resolution Reconstruction of Deformable Tissue from Temporal Sequence of Ultrasound Images

In this paper, the feasibility of super-resolution reconstruction of the deformable tissue from temporal sequences of ultrasound images is extensively studied. The proposed image observation model integrates the non-rigid motion and imaging formation process into a unified super-resolution frame. To facilitate the motion estimation, Lucas-Kanade optical flow is chosen for non-rigid motion estimation from mathematical point of view. Finally, optical flow based iterative back-projection algorithm is proposed for Super-resolution reconstruction, where flow driven diffusion method is developed to re-estimate the motion in the high-resolution coordinate, yielding accurate dense flow field. The efficiency of the proposed approach is demonstrated by simulated images and ultrasound carotid vessel images. Experimental results are provided and compared with existing super-resolution method, which indicate the proposed approach is highly efficient and promising.