Computational geometry provides many solutions to imaging problems, especially for three-dimensional (3D) image compression, segmentation, and measurement. We present here a new method to partition volume data by Voronoi polyhedra structured in a graph environment. A dynamic construction of the 3D Voronoi diagram is proposed, using image information interactively. The process has been applied to segment and quantitate 3D biological data acquired with a confocal laser scanning microscope (CLSM). The discrete volume acquired represents a large mass of data and can be reduced with this particular method, before measurement (processing time) or archiving (memory space). Furthermore, the structure data is a powerful tool to rapidly compute parameters that are characteristic of the volume data.