We examine a methodology for automatically thresholding grayscale images. We begin by suggesting that this endeavor can benefit greatly from a clearly defined "goal". These goals are specific to a given problem, and are often stated in terms of shapes or other spatial characteristics of objects found in the image. The specific techniques that we consider attempt to analyze a criterion function defined in terms of connected-component features from the resultant monochrome images.
P. Kelly, P. Fasel, J. Hogden, J. Howse, and R. Fortson. Towards automated threshold selection for hydrocode simulation imagery. Los Alamos National Laboratory Technical Report LA-UR-06-3899, 2006. [ Abstract | PDF (207 KB) ]
