The recognition systems of patterns in images are mechanisms that filter the information that provides an image to highlight the area of interest for the user. Usually, these mechanisms are based on mathematical transformations that allow the processor to perform interpretations based on the geometry or shape of the image. However, the strategies that implement mathematical transformations are limited, since the effectiveness of these techniques is reduced by changing the morphology or resolution of the image. This paper presents a partial solution to this limitation with a digital image processing technique based on a deep learning neural network (DNN). This technique incorporates a mechanism that allows the DNN to determine the facial expression of a person, based on the segmented information of the image of their face. By segmenting the image and processing its characteristics in parallel, the proposed technique increases the effectiveness of recognizing facial gestures in different images even when modifying their characteristics.