Sensitivity of Wavelet Algorithms in Signal Processing

摘要

During the second year of funding under this grant substantial progress has beenmade in the area of optical flow computations. Following a detailed survey of optical methods for a specially designed set of target tracking problems, a new iterative refinement procedure was developed that compensates for errors in the estimated flow velocity induced by finite difference approximations of spatial and temporal image derivatives. When coupled with a total least squares (TLS) solution scheme (as opposed to a more standard least squares method) and an aggregate velocity assignment for distant targets, this approach yields very accurate flow estimates at a fraction of the computational cost of competing methods. This approach has the pleasant feature that it admits a simple convergence analysis showing that temporal and spatial smoothing improves the iterative convergence for the special case of convective flow. This work has