TY - JOUR
T1 - HOG-NPE: A Novel Local Description Operator for Face Image Recognition
AU - Hongfeng Wang
JO - Journal of Fiber Bioengineering and Informatics
VL - 3
SP - 419
EP - 427
PY - 2014
DA - 2014/07
SN - 7
DO - http://doi.org/10.3993/jfbi09201411
UR - https://global-sci.org/intro/article_detail/jfbi/4797.html
KW - Good Features
KW - Histograms of Oriented Gradient
KW - Subspace Learning
KW - Neighborhood Preserving Embedding
AB - The method of extracting robust feature sets from an image is a crucial issue for the areas of computer
vision and pattern recognition. The nature of real data is a very high dimensional data. However, the
hidden structure can be well characterized by a small number of features in most cases. As a result,
the method of extracting a small number of good features is an important question in computer vision
and pattern recognition, etc. We employ the Histograms of Oriented Gradient (HOG) to extract the
robust feature sets of facial images, which is a local description operator that possesses a certain degree
of invariance against geometric and photometric deformations. Neighborhood Preserving Embedding
(NPE) which is a subspace learning algorithm is adopted to extract a small number of good features on
the local description operators. We use the novel local description operator - HOG-NPE for facial image
recognition, and several experiments on well-known facial databases are conducted, which demonstrate
good performance and effectiveness of this novel local description operator.