The analysis of biometric data is essential in today’s highly connected digital environment. Image processing technology, such as the recognition of identities and facial expression identification, has gained a lot of attention from the academic and technology communities as a result of recent advancements in data collection methods using artificial intelligence and machine learning approaches. Many fields, from surveillance and safety to behavioral science and advertising, can benefit from emotion recognition from facial images. In this research, we propose a novel approach to detecting landmark points that are used as key points of interest (KPOI) on the human face. The suggested method extracts emotions from 2 and 3D images based on landmark points. We evaluated the impact of adding a feature selection approach that prioritizes minimizing redundancy and optimizing relevance on top of the feature extraction process. Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbor (KNN) are three popular classifiers used to uncover the labels. The outcomes validate the efficiency of the feature selection phase. The results suggest that improved performance can be achieved using landmark-based features extracted from 3D images. SVM is the best classifier, followed closely by RF. Using the CK + dataset, researchers found that combining characteristics from 2 and 3D photos yielded the best results. The proposed technique has a higher recognition rate than the current state-of-the-art standard.