Automatic assessment of human femur morphology may provide useful clinical information with regard to hip and knee surgery, prosthesis design and management of hip instability. To this end, neck-shaft and anteversion angles are usually used. We propose a full automatic method to estimate these angles in human femurs. Multislice CT images from 18 dried bones were analysed. The algorithm fits 3D cylinders to different regions of the bone to estimate the angles. A manual segmentation and a conventional angle assessment were used for validation. We found anteversion angle as 20 ± 7° and neck-shaft angle as 130 ± 9°. Mean distances from femur surface to cylinders were 5.5 ± 0.6, 3.5 ± 0.6 and 2.4 ± 0.4 mm for condyles, diaphysis and neck regions, respectively. Automatic and conventional angles were positively correlated (r(2)>0.85). Manual and automatic segmentations did not differ. The method was fast and 100% reproducible. A robust in vivo segmentation algorithm should be integrated to advance towards a clinically compliant methodology.