Determination of hip muscle paths and their moment arms is crucial in subject-specific musculoskeletal modeling. Existing approaches, including via-point, obstacle-set, and others, cannot estimate a smooth path for biarticular muscles such as iliopsoas. This study proposed a modified convex wrapping algorithm based on subject-specific medical imaging to prevent the predicted path from penetrating its underlying bone geometries. Muscle attachment points were automatically personalized using the point cloud registration method. The muscle path was then defined as the shortest path wrapped over the convex hull and searched via Dijkstra’s algorithm. Moreover, a constraint plane for preventing muscle–bone penetration was obtained using the point cloud of the underlying bone. The results of the identified iliopsoas path were compared with the conventional via-point method, revealing that the proposed algorithm can overcome the shortcoming of path non-smoothness via-point paths with certain hip postures. The obtained moment arms were further validated using the magnetic-resonance-imaging-based measurement from the literature, corroborating its applications in musculoskeletal modeling and quantitative surgical planning.