Comparison of Canny and Centroid on Face Recognition Process using Gray Level Cooccurrence Matrix and Probabilistic Neural Network

Abstract

Face recognition system is the development of basic methods of authentication systems by using the natural characteristics of the human face as a basis. The process of recognizing the facial image through several stages of the training phase and testing phase. This study has used datasets in the form of facial image samples obtained with various light intensities, distances, and positions toward the acquisition devices. This study has implemented the Centroid method and Canny edge detection to get image patterns from preprocessed image samples. Image features were obtained from image patterns using Gray Level Co-occurrence Matrix (GLCM). PNN has used as a classification of image patterns. The results of this study showed that the combination of the Centroid and GLCM methods (accuracy of 93.33%) is better than the combination of Canny edge detection and the GLCM method (accuracy of 66.43%). The results of this study also showed that the farther the spatial distance to build the GLCM features, the lower the accuracy.