The precision of 3-D parameters in correspondence-based techniques: the case of uniform translational motion in a rigid environment

A general method of calculating the precision of environmental parameters which differs from conventional methods by finding analytical results, and which does so using only information in the image is discussed. This method called uncertainty-analysis, is applied to the case of first-order-error-based motion algorithms for determining environmental depth by matching corresponding points between two or more frames of a dynamic image sequence. How to calculate the uncertainty of in predicted values for the depth is shown, as well as how to calculate the search region for further instances of the point in subsequent frames, based only on information present in the image. The analysis is applied to a real image sequence with good results.<
>