Estimation of the wetness of objects is an important technique for recognizing states in the real world. In this paper, we propose a non-contact method for estimating the wetness of objects using multispectral near-infrared (NIR) imaging. In contrast with a previous method that requires hyperspectral (110-band) images taken with fine spectral resolution (5 nm intervals) to estimate the degree of wetness, our method enables accurate wetness estimation using few-band NIR images with coarse spectral resolution (40 nm intervals). In general, water absorbs a substantial amount of incident light at wavelengths around 1000 nm and a smaller amount at wavelengths around 900 nm. This phenomenon indicates that the light absorption coefficient of water particularly varies over the NIR spectral band. These differences in the light absorption coefficients of water in the NIR bands are exploited in the model we derived for the appearance of a wet object surface, facilitating accurate wetness estimation. The effectiveness of the proposed method is demonstrated experimentally.