A MAXIMUM-LIKELIHOOD APPROACH TO FEATURE SEGMENTATION

Numerous processes in computer vision are based on the selection of fe atures of interest (e.g. lines) which are then used to infer other inf ormation (e.g. vanishing point coordinates). The selection of features of interest is a difficult problem, first because of inaccuracy of th e feature parameters, and second because of noise caused by the proxim ity of other, unrelated, features. A segmentation method based on the likelihood principle is described in this paper. The reliability of th e process is optimized by using probabilistic models of the parameter uncertainty and of the noise created by the presence of the other feat ures. It is compared with two popular tests; the first based on the Eu clidean distance (called the neighbourhood test), and the second based on the Mahalanobis distance. Then, an example of this method for clas sifying lines with a vanishing point is described and tested on images from indoor scenes. This method has also been successfully applied to other vision segmentation problems (Brillault, A probabilistic approa ch to 3D interpretation of monocular images, Doctoral dissertation, Ci ty University, March (1992)).