Lighting Coefficients Transfer Based Face Illumination Normalization
In this paper, a linear representation based face illumination normalization method is put forward. According to the Lambertian reflectance model, the specific illumination of a face image is determined linearly by a three-dimensional light source orientation vector. Based on the energy function proposed by the Quotient Image approach, we propose a two-step iterative optimization strategy to work out this three-dimensional face illumination representation coefficient vector. Then the neural illumination description coefficients learned from training set are transferred into each face image captured under arbitrary lighting condition to uniform face illumination. Our approach could effectively reduce the disturbance of varying illumination to recognition accuracy. The effectiveness of the proposed method is evaluated on the Extended Yale B database.
Keywordsillumination normalization face recognition Lambertain reflectance surface linear lighting description illumination transfer
Unable to display preview. Download preview PDF.
- 2.Gonzalez, R., Woods, R.: Digital Image Processing, 3rd edn. Prentice Hall, NJ (2006)Google Scholar
- 6.Zhang, L., Yang, M., Feng, X.: Sparse representation or collaborative representation: Which helps face recognition? In: ICCV, pp. 471–478 (2011)Google Scholar
- 10.Xie, X., Lam, K.: Face recognition under varying illumination based on a 2D face shape model. PR 38(2), 221–230 (2005)Google Scholar
- 12.Chen, X., Chen, M., Jin, X., Zhao, Q.: Face illumination transfer through edge-preserving filters. In: CVPR, pp. 281–287 (2011)Google Scholar