Efficient methods for hypothesis verification paradigms

This paper describes efficient methods for verification and their integration in hypothesis verification algorithms suitable for image analysis. The problem of verifying hypotheses can often be reduced (with some restrictions) to the calculation of maximal weighted bipartite graph-matchings or to the calculation of maximal flows minimizing a cost function. These are in known graph theoretic problems and there exist some efficient algorithms for them. These algorithms are faster than graph-searching algorithms, such as A*, and need less memory capacity. Their complexity does not depend on the choice of rating functions or heuristic cost estimation. Experimental results of the integration of such methods and an A*-algorithm as a control-algorithm are included.<>