Quantitative deformable motion estimation for biophysical analysis

A new image sequence analysis algorithm is developed to efficiently calculate the growth rate of plant root cells within the growth zone using video microscopy imaging. The novel approach combines structure tensor and robust matching methods for non-rigid motion-estimation. The plant root is modeled as a volume along its medial axis with non-uniform diameters. After automatic mosaicing of several stacks, the overall growth velocity profile within the root growth zone projected along the root's medial axis is also computed. The robust-tensor algorithm includes several confidence or motion consistency tests to improve accuracy. This method has been verified to be effective in efficiently measuring plant root growth at high spatial and temporal resolution on real roots grown under different experimental conditions. The robust-tensor approach can be extended to measure general biological motion at the cellular level and in other application domains.; In addition to the robust-tensor algorithm, a polynomial-based motion (PMM) model is proposed for the motion estimation and several new optical flow constraints are derived. An algorithm based on this model is proposed and experimental results on standard image sequences and biophysical sequences are presented. The results of this algorithm are more accurate than most existing optical-flow algorithms. This algorithm can be an alternative to the tensor method in the combined tensor and robust matching framework as well as a standalone motion estimation approach. The PMM model leads to a new research area of motion estimation.