Cloud-assisted individual l1-PCA face recognition using wavelet-domain compressed images

Face recognition has been an active research field for a long time, and recently new challenges have arisen in designing cloud-assisted face recognition algorithms. In a cloud assisted face recognition system, mobile devices acquire the data images; then, in order to unbind the cloud face recognition algorithm from the particular features extracted at the mobile device, the images are encoded and uploladed into the cloud. In this framework, it is important to understand and control the effect of the image compression stage performed at the mobile device on the performances of the face recognition algorithms realized within the cloud. Here, we analyze the impact of wavelet domain image compression on the Individual Adaptive (IA) L1-PCA subspace computation and assess the performance of a classifier operating on data characterized by increasing compactness and accordingly decreasing accuracy.