Femur and tibia fractures, are commonly seen in motor vehicle crashes. Cortical bone thickness is an important contributor to bone strength and stress‐strain distribution in impacts. Current finite‐element lower extremity models typically focus on three human sizes (i.e. small female, midsize male, and large male), and do not consider the variation in cortical bone thickness among the whole population. Clinical computed tomography (CT) has been used to determine the cortical bone thickness distribution, but the conventional global thresholding method often fails to offer accurate thickness estimation in thin‐cortex areas. In this study, a new local thresholding method was developed to determine cortical bone thickness using clinical CT scans and statistical models of cortical bone thickness for human femur and tibia were also established with respect to sex, age, stature and body mass index (BMI). It was found that the average thickness error of the newly‐proposed local thresholding method was less than 0.1 mm. In thin‐cortex areas, the proposed method provided more accurate results than the global thresholding method. Statistical analysis suggested that age and BMI significantly affect the cortical bone thickness for both femur and tibia.