Abstract: The low-rank representation (LLR) model can reveal the subtle data structure information and show a strong robustness when dealing with noises．Based on the framework for graph embedding dimensionality reduction method,we proposed a face recognition algorithm which establishes low-rank graph using low-rank representation model．In addition,we constructed a novel low-rank graph with spatial constraint by using spatial information of the tracked points to improve recognition performance．To demonstrate the effectiveness of the presented algorithm,our comparative experiments were conducted using ORL and PIE face image databases.Experimetal results show that the effectiveness and robustness to noises are always better than other state-of-the-art methods.

Key words: Low-rank representation,Spatial constraints,Low-rank graph,Face recognition