<p>Optical density is an important metric for print process control and quality assurance in print production. Conventionally, handheld densitometers are used for this metrology. These devices are expensive and need skilled manpower for operation which may become difficult for small- and medium-scale printing presses. This, in turn, restricts such presses to meet different print standards and business growth, particularly in respect to the global print market. This work proposes the use of a reflective scanner as a possible alternative to such densitometers since scanners are widely available in every press and much less expensive compared to densitometers. For experiments, test targets were printed on a variety of substrates that are commonly used for print production, resulting in a total 3168 color patches. The features were extracted from these patches to build the deep neural network (DNN)-driven prediction model. The model was optimized using various soft computing techniques. The results of different methods were compared and grid search technique has been found most suitable for predicting optical densities from scanned printed patches. The model was compared against some of the existing DNN models and the proposed method has been found to be a promising competitor in terms of prediction accuracy and other standard performance metrics. Based on the considerable performance and consistency, the proposed method can be considered as a feasible, much less expensive and easy to operate alternative of existing densitometer-based process control and evaluation methods in print production.</p>