Wavelet and decision fusion-based 3D face recognition from range image

Abstract

The pivotal purpose of this literature is to describe a new approach for 3D faces recognition in the presence of pose, expression, as well as illumination based on fusion of wavelet coefficients. In addition, authors have investigated the recognition rates with series of experiments by ANN and K-NN. To demonstrate the robustness of the recognition system, Frav3D face dataset has been considered for this investigation. The series of variations in classifiers and their performance accuracies have also ranked using Wilcoxon signed-rank method based on their recognition rates. Range face images from synthesised database are processed by the Haar wavelet transform, and corresponding subimages are created for final fused face dataset. The synthesised database is created by collecting frontal face images along with images obtained after registration of rotated images using ERFI model. Moreover, to discover the features for face recognition, PCA is applied on fused face images. Finally, two supervised classifiers namely, ANN and K-NN are tested for recognition purpose. The obtained maximum recognition rate from our proposed methodology is 96.25% using ANN classifier and 90% of recognition rate from K-NN.