Parameter optimization in the regularized kernel minimum noise fraction transformation 
Allan Aasbjerg Nielsen, Jacob Schack Vestergaard

Abstract  This contribution gives a simple method for finding optimal parameters in a regularized version of the recently suggested kernel minimum noise fraction (kMNF) transformation. The method considers the model signaltonoise ratio (SNR) as a function of the kernel parameter(s) and the regularization parameter. In a grid search we find the parameters that maximize the model SNR. The method is succesfully demonstrated on a remote sensing change detection example with data from the DLR 3K camera system covering a busy motorway. In this example no regularization is chosen and the optimized choice for the kernel parameter gives much better SNR than the default choice. 
Type  Conference paper [With referee] 
Conference  IEEE IGARSS 
Year  2012 Month July pp. 370373 
Address  Munich, Germany 
Electronic version(s)  [pdf] 
BibTeX data  [bibtex] 
IMM Group(s)  Image Analysis & Computer Graphics, Geoinformatics 