Optical density (OD) is a crucial control parameter during the printing process. It is typically measured using densitometers, which are precise but cost-prohibitive for widespread use. A novel method is proposed for OD measurement using scanners and a stacked ensemble learning model with a neural network meta-model. The proposed method has shown promise in overcoming some of the inherent limitations of the scanner itself, such as the non-colorimetric sensors and nonlinear responses. A dataset of 864 patches covering three common substrate types was prepared. Advanced feature engineering methods were implemented. The stacked ensemble model, consisting of many different base regressors along with a neural network meta-model, achieved promising results with R2 values up to 0.9787 for the training data and 0.9599 for the testing data. The presented method outperforms existing attempts reported in the literature by more than 61%. Outcomes suggest that the proposed method can serve as an affordable alternative to traditional densitometric measurement with adequate real-time print process control capability. This work demonstrates the potential that an inexpensive imaging device integrated with machine learning can help democratize metrological capabilities in the printing industry.