Enhanced Local Texture Feature Sets for Face Recognition Under Difficult Lighting Conditions IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 19, NO. 6, JUNE 2010 Xiaoyang Tan and Bill Triggs 報告者:王克勤 1 Introduction • Face recognition has received a great deal of attention from the scientific and industrial communities over the past several decades • This paper focuses mainly on the issue of robustness to lighting variations 2 Traditional approaches • Appearance-based • Normalization-based • Feature-based 3 Appearance-based approaches • Training examples are collected under different lighting conditions and directly used to learn a global model of the possible illumination variations • Direct learning of this kind makes few assumptions but it requires a large number of training images and an expressive feature set 4 Normalization-based approaches • Normalization based approaches seek to reduce the image to a more canonical form in which the illumination variations are suppressed • Histogram equalization 5 Histogram equalization • A method in image processing of contrast adjustment using the image's histogram http://en.wikipedia.org/wiki/Histogram_equalization 6 Feature-based approaches • Feature-based approaches extracts illumination-insensitive feature sets directly from the given image • Local binary patterns(LBP) 7 Local binary patterns(cont.) 1 1 1 1 0 1 1 0 8 0 32 2 4 16 64 128 LBP=1X1 + 1X2 + 1X4 + 1X8 + 1X32 =47 10 11 • Appearance-based approaches • Normalization-based approaches • Feature-based approaches 12 • Preprocessing chain • LTP local texture feature sets • Multiple-feature fusion framework 13 Preprocessing chain 14 Gamma correction • Gamma correction is a nonlinear gray-level transformation • Replace gray-level with or (for ) 15 Difference of Gaussian Filtering • Gamma correction does not remove the influence of overall intensity gradients such as shading effects • High-pass filtering removes both the useful and the incidental information 16 Difference of Gaussian Filtering(cont.) • Difference of Gaussians is a grayscale image enhancement algorithm that involves the subtraction of one blurred version of an original grayscale image from another, less blurred version of the original • Difference of Gaussians can be utilized to increase the visibility of edges and other detail present in a digital image http://en.wikipedia.org/wiki/Difference_of_Gaussians 17 Masking • If facial regions (hair style, beard, ) that are felt to be irrelevant or too variable need to be masked out, the mask should be applied at this point 18 Contrast equalization • This stage rescales the image intensities to standardize a robust measure of overall contrast or intensity variation 19 Local ternary patterns • Local binary patterns threshold at exactly the value of the central pixel tend to be sensitive to noise • This section extends LBP to 3-valued codes, LTP 20 Local ternary patterns(cont.) The tolerance interval is [49, 59] 21 Local ternary patterns(cont.) 22 Local ternary patterns(cont.) 23