An illumination tolerant class specific 2D subspace based face recognition technique using optimum correlation filter

作者：Banerjee Pradipta K; Datta Asit K^*

来源：Optik, 2013, 124(17): 3173-3179.

DOI：10.1016/j.ijleo.2012.09.053

• Correlation filter
• Subspace
• Face recognition
• Correlator

This paper presents a technique of frequency domain correlation for illumination tolerant face recognition. A pair of correlation filters are generated from input image projected onto class specific subspace, called optimum projecting image correlation filter and from reconstructed image, called reconstructed correlation filter. As the test face image dynamically configures the nature of these two filters, illumination tolerance is achieved. A sharp peak at the correlation plane measured by peak to sidelobe ratio indicates the detection of test face image. The performance of the proposed strategy is evaluated on YaleB and PIE database. The proposed technique outperforms the other existing correlation filters. A hybrid optical system is used for the hardware implementation of the proposed system.

• 出版日期2013