Precision estimation of a fitted image line

A simple and systematic treatment of the problem of estimating the precision of a fitted image is discussed. It is taken as an error propagation problem, and the general jacobian linearization approach is used to analyze it. Then, the formulas of the estimated precision of the fitted line are derived for two popular line fitting approaches: one is the least-squares line fitting, and the other is the least eigenvalue line fitting, under two typical noise models. To evaluate the obtained precision formulas, some independent Monte-Carlo simulations are made to judge the accuracy and robustness, and then some comparisons with some existing results from complex approaches are made. Several experimental tests in real systems are also mentioned.<
>