A Low Complexity Face Recognition System using Discrete Wavelet Transform and Distance Metrics

Abstract

Face recognition (FR) is a biometric technology used to identify persons according to their facial features. This study proposes an FR system that meets the proper design requirements, including high accuracy, low complexity, and high processing speed. The multi-level Discrete Wavelet Transform (DWT) is used to extract reduced-size faces' features, whereas the recognition is accomplished using the Euclidean distance. The maximal overlap DWT (MODWT) is used as a feature extractor, which is applied to the obtained features of DWT. The experiments show that the results of using various levels of DWT, without and with the MODWT, are somewhat close to each other using the ORL, Yale, and Face95 databases. In comparison, the proposed work surpasses other related approaches; it highly reduced the utilized data, resulting in a small memory occupation, high processing speed, and a high accuracy rate.