The aim of this in vitro study was to evaluate the ability of a dental subtraction radiography system to quantitatively detect differences in density between radiographic image pairs. Four periapical radiographs were taken of the upper first permanent molars on five human skulls using the Digora radiographic imaging system. The 4 images were a "baseline" image and 3 containing test objects consisting of either 0.5, 1 or 2 mm thick aluminium cylinders, 2.5 mm in diameter. Semi-automated image processing software was used to "warp" the 3 images with test objects into the same geometric/density registration as the corresponding baseline image using a process called patch minimization. "Difference" images were than produced and their contrast stretched. For regions of interest, with and without test objects present, the difference in density between the baseline and "test object" images was calculated using a reference aluminium step wedge. The test objects were clearly visible in all the "difference" images. The mean difference between the actual and estimated volume of the test object was 0.31 (95% CI [-0.55, 1.17]) mm3 Al. There was a strong association (r = 0.83) between the actual and estimated aluminium volumes. It is concluded that this system provides adequate precision for clinical evaluation.