Osteoporosis, a bone fragility disease leading to spontaneous bone fractures, is becoming a major health problem in many countries. The characterization of trabecular bone architecture has been shown to be important to predict fracture risk. Trabecular bone may be characterized in vitro using 3D microtomography. New acquisition systems enable to get very high-resolution images with sizes up to 8 Gbytes. Among the different quantifications, connected component analysis is useful to characterize connectivity. Such an analysis requires loading the whole 3D image in memory, which limits the maximum size of the volume to be processed. In this paper, we present a method for analyzing the connectivity of very large 3D images. They are decomposed in two or more sub-volumes and processed according to the following steps: a) objects in each sub-volume are separately labeled by an iterative sequential algorithm, b) the common border of each pair of sub-volumes is analyzed in order to identify interconnections between objects situated on both sides of the border, c) objects in each sub-volume are relabeled according to interconnections information. The implementation is thought so as to minimize memory load. The program was successfully applied to bone volume images acquired using synchrotron radiation microtomography and containing various numbers of objects