Local histogram specification for face recognition under varying lighting conditions |
| |
Authors: | Hui-Dong Liu Ming Yang Yang Gao Chunyan Cui |
| |
Affiliation: | 1. School of Computer Science and Technology, Nanjing Normal University, Nanjing 210023, PR China;2. State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, PR China |
| |
Abstract: | High frequency illumination and low frequency face features bring difficulties for most of the state-of-the-art face image preprocessors. In this paper, we propose two methods based on Local Histogram Specification (LHS) to preprocess face images under varying lighting conditions. The proposed methods are able to significantly remove both the low and high frequency parts of illumination on face images, as well as enhance face features lying in the low frequency part. Specifically, we first apply a high-pass filter on a face image to filter the low frequency illumination. Then, local histograms and local histogram statistics are learned from normal lighting images. In our first method, LHS is applied on the entire image. By contrast, in the second method, the regions contain high frequency illumination and weak face features on a face image are identified by local histogram statistics, before LHS is applied on these regions to eliminate high frequency illumination and enhance weak face features. Experimental results on the CMU PIE, Extended Yale B and CAS-PEAL-R1 databases demonstrate the effectiveness and efficiency of our methods. |
| |
Keywords: | Local histogram specification Local histogram statistics Face recognition High frequency illumination Low frequency face features |
本文献已被 ScienceDirect 等数据库收录! |
|