Background:
Age estimation using face images has become increasingly significant in
the recent years, due to diversity of potentially useful applications. Age group feature extraction, the
local features, has received a great deal of attention.
Objective:
This paper derived a new age estimation operator called “Gradient Dual-Complete Motif
Matrix (GD-CMM)” on the 3 x 3 neighborhood of gradient image. The GD-CMM divides the 3 x 3
neighborhood in to dual grids of size 2 x 2 each and on each 2 x 2 grid complete motif matrices are
derived.
Methods:
The local features are extracted by using Motif Co-occurrence Matrix (MCM) and it is derived
on 2 x 2 grid and the main disadvantage of this Motifs or Peano Scan Motifs (PSM) is they are
static i.e. the initial position on a 2 x2 grid is fixed in deriving motifs, resulting with six different motifs.
The advantage 3 x 3 neighborhood approaches over 2x 2 grids is the 3x3 grid identify the spatial
relations among the pixels more precisely. The gradient images represent facial features more efficiently
and human beings are more sensitive to gradient changes than original grey level intensities.
Results:
The proposed method is compared with other existing methods on FGNET, Google and
scanned facial image databases. The experimental outcomes exhibited the superiority of proposed
method than existing methods.
Conclusion:
On the GD-CMM, this paper derived co-occurrence features and machine learning
classifiers are used for age group classification.