Bilinear Feature Line Analysis for Face Recognition

A novel Bilinear Feature Line Analysis (BFLA) is proposed for image feature extraction in this letter. Neaerest feature line (NFL) is a powerful classifier. Some NFL based subspace algorithms have been introduced recently. In most of the classical NFL-based subspace learning approaches, the input samples are vectors. For face recognition, face samples should be transformed to vectors firstly. This process induces a high computational complexity and also may lead to the loss of the geometric feature of samples. The proposed BFLA is a matrix-based algorithm. It aims to minimize the within class scatter based on two-dimensional NFL. The experimental results on Yale face databases confirm its effectiveness.