Recognition of Face Images with Noise Based on Tucker Decomposition
Lukáš Zaorálek, Michal Prílepok, Václav Snåšel · 2015
The main goal of this paper is to detect faces from noisy images using three different classification methods and compare the results obtained from the classification methods. The faces are described by a set of images. Many other unsupervised statistical algorithms such as Principal Component Analysis (PCA) or Singular Value Decomposition (SVD) use only one image per person to extract features from the face. These approaches can lose important information, for example a relationship between images of the same person taken under different conditions. It shows that data structure like tensor and it decomposition increase the quality of recognition in this task because it better captures important features of one face taken from several images. The accuracy of the tensor approach is compared with other well-known techniques such as Support Vector Machine (SVM) and Neural Network (NN).