Purpose: To assess the reliability of cone beam computed tomography (CBCT) voxel gray value measurements using Hounsfield units (HU) derived from multislice computed tomography (MSCT) as a clinical reference (gold standard). Materials and Methods: Ten partially edentulous human mandibular cadavers were scanned by two types of computed tomography (CT) modalities: multislice CT and cone beam CT. On MSCT scans, eight regions of interest (ROI) designating the site for preoperative implant placement were selected in each mandible. The datasets from both CT systems were matched using a three-dimensional (3D) registration algorithm. The mean voxel gray values of the region around the implant sites were compared between MSCT and CBCT. Results: Significant differences between the mean gray values obtained by CBCT and HU by MSCT were found. In all the selected ROIs, CBCT showed higher mean values than MSCT. A strong correlation (R = 0.968) between mean voxel gray values of CBCT and mean HU of MSCT was determined. Conclusions: Voxel gray values from CBCT deviate from actual HU units. However, a strong linear correlation exists, which may permit deriving actual HU units from CBCT using linear regression models. Keywords: accuracy, CBCT, cone beam computed tomography, Hounsfield unit