Image enhancement techniques are discussed in this paper as a necessary preprocessing step for face detection. First, a measure of the distribution of image information, termed the entropy error rate (EER), is presented on the basis of information theory. Then, by integrating a histogram ridge analysis technique and an optimal intensity transform method that aims to minimize the EER of an enhanced image, a novel adaptive enhancement algorithm is proposed. In a baseline face detection test using the algorithm presented by Viola et al. in (2001), comparison experiments are conducted with the Yale B face dataset and our own movie face dataset. The results demonstrate that image enhancement preprocessing can significantly improve face detection accuracy, and that the adaptive enhancement algorithm performs much better than classical histogram-based enhancement techniques such as linear stretching and histogram equalization.
[1]
Paul A. Viola,et al.
Robust Real-Time Face Detection
,
2001,
International Journal of Computer Vision.
[2]
Azriel Rosenfeld,et al.
Face recognition: A literature survey
,
2003,
CSUR.
[3]
Erik Hjelmås,et al.
Face Detection: A Survey
,
2001,
Comput. Vis. Image Underst..
[4]
David J. Kriegman,et al.
From Few to Many: Illumination Cone Models for Face Recognition under Variable Lighting and Pose
,
2001,
IEEE Trans. Pattern Anal. Mach. Intell..
[5]
Narendra Ahuja,et al.
Detecting Faces in Images: A Survey
,
2002,
IEEE Trans. Pattern Anal. Mach. Intell..
[6]
Du-Ming Tsai,et al.
A fast histogram-clustering approach for multi-level thresholding
,
1992,
Pattern Recognit. Lett..