Evaluation of Illumination Compensation Approaches for ELGBPHS
Various environmental conditions like pose variations, scale, noise and illumination changes cause matching problems for face recognition algorithms due to the fact that inappropriate data from images is extracted and consequently the recognition rate suffers. In the worst case, persons who should be accepted are rejected and vice versa. Enhanced Local Gabor Binary Patterns Histogram Sequence (ELGBPHS) is considered as an advanced and robust face recognition method. In this paper we evaluated if state-of-the-art illumination compensation approaches can further improve the performance of ELGBPHS. The paper outlines if it is worth to additionally implement preprocessing steps with the increasing complexity and cost. Therefore tests were performed to check if the recognition rate improves if applying preprocessing steps and adjusting essential parameters. Multi-Scale-Retinex, Histogram Equalization, 2D discrete Wavelet-Transformation and one approach combining Gamma Correction, Difference of Gaussian Filtering and Contrast Equalization (TT) were implemented and evaluated.
KeywordsFace Recognition Recognition Rate Face Image Local Binary Pattern Face Detection
Unable to display preview. Download preview PDF.
- 1.Acharya, T., Ray, A.K.: Image Processing: Principles and Applications. Wiley-Interscience, Hoboken (2005)Google Scholar
- 3.Bolme, D.: Elastic bunch graph matching. Master’s thesis, Colorado State University (2003)Google Scholar
- 4.Du, S., Ward, R.: Wavelet-based illumination normalization for face recognition. In: Proceedings of the 2005 IEEE International Conference on Image Processing (ICIP 2005), pp. 954–957. IEEE Computer Society, Los Alamitos (2005)Google Scholar
- 6.Intel. Opencv library (2010), http://opencv.willowgarage.com/wiki/