Clustering and estimation of 2-D motion fields in ultrasonic images based on regression characteristics with respiratory signal

摘要

We estimated instantaneous 2-D motion field in ultrasonic images of internal organs. We have proposed an algorithm based on the correlation between respiration and the motion of internal organs, in which, motion is represented as a regression model with respect to the respiratory signal and independently estimated in each pixel. In this paper, it is assumed that motion has a multi-modal distribution, thereby reducing the number of unknown parameters, and hence improving the accuracy of the motion estimated. We constructed a computationally stable estimation algorithm based on this assumption, and verified the applicability of the proposed technique through experiments on simulated and real image data.