Novel illumination-normalization method for Face Recognition
Ching Chung Huang and Cheng Yi Liu
The flow chart of illumination-normalization algorithm.
The block diagram of a face recognition system.
Illumination normalization is an important preprocessing step for many operations such as face recognition. Therefore, a suitable illumination-normalization is quite useful and plays an important key technology for those applications. It is well known that the image gray level (or image color) is very sensitive to the lighting variation. The same object with different illumination may produce considerably different images. For machine vision system, it is difficult to produce good classification accuracy if image samples in the training and testing sets are taken under different lighting conditions. In this project, the goal is to make the testing face image insensitive to the lighting environment or modify the testing image with similar illumination information.

This work is supported by Industrial Technology Research Institute (Advanced Technology Center) in Taiwan.
Some results after the illumination-normalization mechanism. (a) The original images (b) The face images (c) The images after illumination-normalization.
The recognition result using different illumination-normalization mechanism in MIT face database. Method A means no illumination normalization is used. Method B uses histogram equalization as illumination normalization. Method C uses the method proposed in [9]. Method D uses the proposed method.
The recognition result using different illumination-normalization mechanism on ATC database. The computation time means the average time used to normalize a testing image.
Ching-Chun Huang and Cheng Yi Liu "Novel Illumination-Normalization Method Based on Region Information", Proceedings of SPIE, San Jose, USA, March 2005, pp. 339-348. (EI).