Hybrid techniques to improve face recognition based on features extraction methods and Haar discrete wavelet transformation

Abstract

This paper uses hybrid techniques to improve the rate of recognition for a face from identified data set of faces. These techniques are summarized by applying firstly the Haar discrete wavelet transformation method in order to enhance and compress the images of the data set and store the results for each process in a separate data set. Secondly, applying a hybrid method from two popular face recognition methods called Principal Component Analysis and Singular Value Decomposition for extracting feature from the images. This work applied by using a dataset contains 400 images for 40 different persons called Olivetti Research Laboratory (ORL). In calculating the distances between image vectors, Manhattan measurement is used and its show a very good results in recognition rate. From this work, it can be concluded that recognition rate increments with the decrementing in the number of dataset images and increasing the threshold value. The expended time in execution decreases in a very obvious way when using the compressed dataset rather than the enhanced dataset which its images has four times the size of the images in the compressed dataset.