To understand the kinematic effects of surgery, arthroplasty or conservative treatments, a noninvasive system to capture accurate 3D imaging of functional activities in prospective, controlled studies is required. To provide such a technique, a new algorithm was developed to register 3D CT data of normal bones to the same bones in a 2D fluoroscopy frame. The algorithm produces a digitally reconstructed radiograph (DRR) from the CT data and then filters this to produce an edge-enhanced image. The resulting image is then registered with an edge-enhanced version of the fluoroscopy frame using a new similarity measure called Cross-Correlation Residual Entropy (CCRE). The system was evaluated by implanting tantalum beads into three cadaveric knees to act as fiducial markers. The knees were flexed between 0 degrees and 70 degrees , and single-plane fluoroscopy data of the knees were acquired. CT data of the femur and tibia were then individually registered to the fluoroscopy images. No significant measurement bias was observed, and the standard deviation of the error in bead positions was 0.38 mm for in-plane translation and 0.42 degrees for rotation. To determine the accuracy of the registration algorithm for out-of-plane translations, fluoroscopy frames were scaled in size by fixed increments; the average standard deviation of the errors for out-of-plane translation was 0.65 mm. The ability to obtain such accurate 3D motion data from a noninvasive technique will enable prospective, longitudinal, and controlled studies of reconstruction surgery, and conservative management of joint pathologies.