Use this URL to cite or link to this record in EThOS:
Title: Multi-scale Local Binary Pattern histogram for face recognition
Author: Chan, Chi Ho
ISNI:       0000 0001 3526 5971
Awarding Body: University of Surrey
Current Institution: University of Surrey
Date of Award: 2008
Availability of Full Text:
Access from EThOS:
Access from Institution:
Recently, the research in face recognition has focused on developing a face representation that is capable of capturing the relevant information in a manner which is invariant to facial expression and illumination. Motivated by a simple but powerful texture descriptor, called Local Binary Pattern (LBP), our proposed system extends this descriptor to evoke multiresolution and multispectral analysis for face recognition. The first descriptor, namely Multi-scale Local Binary Pattern Histogram (MLBPH), provides a robust system which is relatively insensitive to localisation errors because it benefits from the multiresolution information captured from the regional histogram. The second proposed descriptor, namely Multispectral Local Binary Pattern Histogram (MSLBP), captures the mutual relationships between neighbours at pixel level from each spectral channel. By measuring the spatial correlation between spectra, we expect to achieve higher recognition rate. The resulting LBP methods provide input to LDA and various classifier fusion methods for face recognition. These systems are implemented and compared with existing Local Binary Pattern face recognition systems and other state of art systems on Feret, XM2VTS and FRGC 2.0 databases, giving very promising results in the controlled environment. Photometric normalisation is important for face recognition, even if illumination-robust features, such as Gabor or LBP, are used for face representation. In order to study the merits of photometric normalisation, five different photometric normalisation methods have been investigated. A superior performance is achieved by MLBPH with the Preprocessing Sequence method in all the tests. The results of a comparison with the state-of-art systems show that the proposed Multi-scale Local Binary Pattern histogram method with the Preprocessing Sequence photometric normalisation achieves similar performance to the best performing systems, its key advantage is that it offers a simple solution which is robust to localisation errors and changing illumination.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available