Face super-resolution using a hybrid model
Liu Li, Yiding Wang · 2008
Face super-resolution is to synthesize a high-resolution facial image from a low-resolution input, which can significantly improve the recognition for computer and human. In this paper, we propose a new method of super-resolution based on hybrid model including a linear model of eigenface super-resolution and a Bayesian formulation model. Principal Component Analysis (PCA) is used to approximately represent the input face image by linear combination of limited eigenface images. Then preliminary estimation of super-resolution result can be given by hallucinating the low-resolution eigenface images in the linear combination representation respectively. Finally, we use a Bayesian estimation algorithm to consider of the effect brought by subspace representation error and observation noise. Our method is demonstrated by extensive experiments with promising results of high-quality hallucinated results.