Recent advances in biomedical imaging technology have made possible the visualization of various objects ranging from internal organs to microorganisms. Quantitative analysis of these image data provide essential information for clinical and biological research. In this dissertation, we present algorithms for the identification, segmentation and analysis of biomedical objects in several problems, most of which pose difficulties because of their high dimension, high quantity and high complexity. The problems we studied include objects with complicated branching and networking structure such as the double walls of human airway trees and fibrin networks in blood clots, as well as small objects with massive quantities or temporal behaviors such as the cells in lymph node tissue sections and moving bacteria. The proposed algorithms are based on combinations of a variety of techniques including optimal graph search, medial axis analysis, graph matching, shape feature extraction and machine learning. Our approaches either provide tools for automatic and reliable analysis, or significantly reduce the human effort involved in interactive analysis. As a result, we also perform some quantitative analysis based on the output of our algorithms, which generate biologically or diagnostically meaningful findings.