Global Optimization in Computer Vision

In this paper, we develop an algorithm for minimizing the $L_{p} $ norm of a vector whose components are linear Ractional fimctions, where $p $ is an arbitraly positive integer. The problem is included in the sum-of-ratios problem, and often occurs in computer vision. In that case, it is characterized by a large number of ratios and a small number of vari-ables. The algorithm we propose exploits those special stmctures and generates a globally optimal solution in a practical amount of computational time