A method for selection of optimal features Is developed as part of the solution of a multi-alter-naltivs problem. Hie problem is solved in three stages. At the first stage, an initial multlcom-pouent image is converted to an intermediate image with few components. In the second stage, the recognition proper takes place. The third stage is carried out for the further refinement and can be included in the presence of pairwSse groupabk images. The optimum criteria for two features are defined. Experimental results are reported.
展开▼