Face recognition is one of the interesting and emerging research area in many security domains. Still, the accurate recognizing of face parts is a complicated task due to their different poses, illumination effects, expressions and artifacts. So, the existing works used conventional machine learning techniques for improving the detection rate and accuracy of face recognition systems. But it limits with the issues related to high complexity, reduced efficiency, mis-classified results, and increased error rate. In order to solve these issues, this work intends to develop an enhanced deep learning model for a competent face recognition system. Here, three different types of datasets like MUCT, ORL, and Grimace are used for testing the face recognition system.At first, the geometrical face parts detection mechanism is implemented to crop the face region from the given input. Then, the Guided Face Filtering (GFF) technique is employed to reduce the noise artifacts and to improve the quality of image based on the histogram equalization. The most suited features used for the classification are extracted from the normalized face image by using the Convolutional Neural Network (CNN) model. During experimentation, the performance of the proposed GFF-CNN is validated and compared by using various evaluation metrics.

Face Recognition, Deep Learning, Convolutional Neural Network (CNN), Filtering, Feature Selection Geometrical Face Parts Detection, and Guided Face Filtering (GFF).