Evaluation of comparison levels in iterative estimators

In image analysis two stages can be considered, object recognition and object parameter estimation. An initial estimate of the object parameters can be generated in the object recognition stage. The paper is about the second stage. The paper shows the usability of different comparison levels in iterative estimators. The levels of comparison studied are raw pixel data, shape of segments and shape parameters of these segments. It is demonstrated that accuracies as required by vision problems encountered in robot vision can be achieved at the different comparison levels.<
>