The uptake of moisture severely affects the properties of wood in service applications. Even local moisture content variations may be critical, but such variations are typically not detected by traditional methods to quantify the moisture content of the wood. In this study, we used near-infrared hyperspectral imaging to predict the moisture distribution on wood surfaces at the macroscale. A broad range of wood moisture contents were generated by controlling the acetylation degree of wood and the relative humidity during sample conditioning. Near-infrared image spectra were then measured from the surfaces of the conditioned wood samples, and a principal component analysis was applied to separate the useful chemical information from the spectral data. Moreover, a partial least squares regression model was developed to predict moisture content on the wood surfaces. The results show that hyperspectral near-infrared image regression can accurately predict the variations in moisture content across wood surfaces. In addition to sample-to-sample variation in moisture content, our results also revealed differences in the moisture content between earlywood and latewood in acetylated wood. This was in line with our recent studies where we found that thin-walled earlywood cells are acetylated faster than the thicker latewood cells, which decreases the moisture uptake during the conditioning. Dynamic vapor sorption isotherms validated the differences in moisture content within earlywood and latewood cells. Overall, our results demonstrate the capabilities of hyperspectral imaging for process analytics in the modern wood industry. Graphical abstract: [Figure not available: see fulltext.].