Markowitz Model for Robust Edge Detection

For many image processing applications, the aim is to detect a certain event or occurrence in a noisy data set. Several algorithms may exist that solve the detection problem. An example is the detection of edges. The subsequent difficulty then is how to select a proper weighting scheme for the algorithms so that the results are combined optimally. To achieve proper fusion of detection outputs, in this article, the weighting scheme is based on the Markowitz model derived from the principles of diversification. We verify, in theory and in practice, that combining detection outputs based on the Markowitz model significantly improves the error-standard deviation combinations for non-correlated edge detection algorithms.