Noise reduction, restoration, and segmentation methods are developed for the quantitative structural analysis in three dimensions of aggregated oil-in-water emulsion systems imaged by fluorescence confocal laser scanning microscopy. Mindful of typical industrial formulations, the methods are demonstrated for concentrated (30% volume fraction) and polydisperse emulsions. Following a regularized deconvolution Step using an analytic optical transfer function and appropriate binary thresholding, novel application of the Euclidean distance map provides effective discrimination of closely Clustered emulsion droplets with size variation over at least 1 order of magnitude. The a priori assumption of spherical nonintersecting objects provides crucial information to combat the ill-posed inverse problem presented by locating individual particles. Position coordinates and size estimates are recovered with sufficient precision to permit quantitative study of static geometrical features. In particular, aggregate morphology is characterized by a novel void distribution measure based on the -eneralized Apollonius problem. This is also compared with conventional Voronoi/Delauney analysis.