3
Figure 1: Proposed multi-level face recognition taxonomy.
1. Face structure – This level relates to the way a
recognition solution deals with the facial structure,
considering three classes: i) global representation,
dealing with the face as a whole (see Figure 2.a);
ii) component + structure representation, relying only
on the characteristics of some face components, such as
eyes, nose, mouth, etc., along with their relations (see
Figure 2.b); and iii) component representation, dealing
independently with a meaningful selection of face
components, without any consideration of the relations
between them (see Figure 2.c).
Figure 2: Face structure level: (a) global; (b) component +
structure; and (c) component representation face structures.
2. Feature support – This level is related to the (spatial)
region of support considered for feature extraction,
which can be either global or local. Global feature
support implies that all the selected facial structure area
is considered as region of support for feature extraction,
corresponding to either the full face (Figure 3.a) or a full
face component (Figure 3.b), without any further
partitioning. On the contrary, local feature support
implies that the region of support for feature extraction
is a smaller part of either the full face (Figure 3.c) or the
face component (Figure 3.d). A local region of support
can have different characteristics, for instance in terms of
topology, size, overlapping, among others; a much used
type of partitioning is the simple square based division
of the face or component.
Figure 3: Feature support level: Global feature support with
(a) global and (b) component face structures; Local feature
support with (c) global and (d) component face structures.
The square blocks in (c) and (d) represent the local (spatial)
support considered for feature extraction.
3. Feature extraction approach – This level is related to
the specific feature extraction approach, which may be
classified as: i) appearance based, deriving features by
using statistical transformations from the intensity data;
ii) model based, deriving features based on geometrical
characteristics of the face; iii) learning based, deriving
features by modelling and learning relationships from the
input data; and iv) hand-crafted based, deriving features
from pre-selected elementary characteristics.
4. Feature extraction sub-approach – The last level
considered in the proposed taxonomy is a sub-category
of the previous one, to allow identifying the specific
family of techniques used by the selected feature
extraction approach.
Regarding the feature extraction approach and sub-
approach levels, these are directly related to the specific
feature description technology adopted, as discussed in the
following.