The high transparency of dental enamel in the near-infrared (NIR) at can be exploited for imaging dental caries without the use of ionizing radiation. The objective of this study is to determine whether the lesion contrast derived from NIR imaging in both transmission and reflectance can be used to estimate lesion severity. Two NIR imaging detector technologies are investigated: a new Ge-enhanced complementary metal-oxide-semiconductor (CMOS)-based NIR imaging camera, and an InGaAs focal plane array (FPA). Natural occlusal caries lesions are imaged with both cameras at , and the image contrast between sound and carious regions is calculated. After NIR imaging, teeth are sectioned and examined using polarized light microscopy (PLM) and transverse microradiography (TMR) to determine lesion severity. Lesions are then classified into four categories according to lesion severity. Lesion contrast increases significantly with lesion severity for both cameras . The Ge-enhanced CMOS camera equipped with the larger array and smaller pixels yields higher contrast values compared with the smaller InGaAs FPA . Results demonstrate that NIR lesion contrast can be used to estimate lesion severity.