Detection and Segmentation of Human Faces
in Color Images with Complex Backgrounds
Manika Singhal#1,Somya Garg#2,Nidhi Garg#3
Department of CSE, Raj Kumar Goel Institute of Technology for Women
Ghaziabad, India
1manikasinghal241@gmail.com
2mastsomya@gmail.com
3nidhigarg@rkgitw.edu.in
Abstract-This paper proposes a novel fast approach to the
detection, segmentation and localization of human faces in
color images under complex background. First, a number of
evolutionary agents are uniformly distributed in the 2-D
image environment to detect the skin-like pixels and segment
each face-like region by activating their evolutionary
behaviors. Then wavelet decomposition is applied to each
region to detect the possible facial features and a three-layer
BP neural network is used to detect the eyes among the
features. Experimental results show that the proposed
approach is fast and has a high detection rate.
morphological operations is used for binary images. After
this fill image regions and holes based on morphological
reconstruction. Since this stage aims to find out to exact
face. Thus a morphological close and open algorithms are
applied on the resultant reconstructed image. Thus at last
the morphological dilation with a square Structuring
element is used to eliminate the effects of slightly slanted
edges and a vertical linear structuring element is then
employed in a closing operator to force the strong vertical
edges clogged.
III.
Keywords-Face detection; Face location; Image
segmentation; Evolutionary computation; Agent;
Wavelet decomposition; Neural network.
I.
INTRODUCTION
One of the most popular topics of research in the
computer vision field. It has many applications including
face recognition, crowd surveillance, and human
computer interaction. Face detection is to locate and
detect all the faces and size from input images. It is
complicated because of complex background, various
illumination conditions, and large amount of facial
expressions and so on. Therefore we have to find an
effective approach which is somehow invariant to these
changes.
It may be felt that face detection is a trivial task. After all,
we human beings, do this in our daily lives without any
effort. The human visual system can easily detect and
differentiate a human face from its surroundings but it is
not easy to train a computer to do so. The detection of a
face in a scene comes out to be an easy task for the
human brain due to a lot of parallelism achieved by it. It
has not been possible for the present day computational
systems to reach that level of parallelism in processing
information.
II.
FACE DETECTION
In this step, the morphologically operation with the
structuring element apply in the image and generally the
FACE LOCALIZATION
When human face candidate area screening, as a result of
the forehead part superposition, as well as the forehead with
other parts, for example clothes and so on the connection,
to screened has caused the difficulty, therefore should use
first shuts the operation, separated the connection, then
carried on processing. So in this step, segment out non-face
regions using major to minor axis ratio with help of
Heuristic Filtering. Only those regions in the retained
image which have an area greater than or equal to 1/250 of
the maximum area region and remove those regions which
have Width/Height < 6 ratio.
IV.
FACE FEATURE CLASSIFICATION
The purpose of this section is to bound boundary box the
faces from the localization text region. As the main color in
a human faces is skin color, the threshold should be big
enough. On the other hand, face region contains eyes,
mouth, hair, nose and other background colors, therefore
the threshold cannot be too high.
V.
FACE DETECTION ALGORITHM
Face detection contains two major modules:
(i) Face localization for finding face candidates and
(ii) Facial feature detection for verifying detected face
candidates.
The eyes and mouth candidates are verified by checking
(i) Luma variations of eye and mouth blobs;
(ii) Geometry and orientation constraints of eyes-mouth
triangles; and
(iii) The presence of a face boundary around eyes-mouth
triangles.
Fig.-1 Face Detection
A. Lighting compensation and skin tone detection
The appearance of the skin-tone color can change due
to different lighting conditions.
We introduce a
lighting compensation technique that uses “reference
white” to normalize the color appearance. We regard
pixels with top 5 percent of the luma (nonlinear
gamma-corrected luminance) values as the reference
white if the number of these reference-white pixels is
larger than 100. The R, G and B components of a
color image are also adjusted in the same way as these
reference-white pixels are scaled to the gray level of 255.
Based on the locations of eyes/mouth candidates, our
algorithm first constructs a face boundary map from the
luma, and then utilizes a Hough transform to extract the
best-fitting ellipse. The fitted ellipse is associated with a
quality measurement for computing eyes-and-mouth
triangle weights. An ellipse in a plane has five
parameters: an orientation angle, two coordinates of the
center, and lengths of major and minor axes. Since we
know the locations of eyes and mouth, the orientation
of the ellipse can be estimated from the direction of a
vector that starts from midpoint between eyes to the
mouth. The location of the ellipse center is estimated from
the face boundary. Hence, we only require a twodimensional accumulator for an ellipse in a plane. The
ellipse with the highest vote is selected. Each eye-mouth
triangle candidate is associated with a weight that is
computed from its eyes/mouth maps, ellipse vote and
face orientation that favors vertical faces and symmetric
facial geometry.
Fig.2 Eample of Face Detection Technique
Fig.3 Localizatipn of Facial Features
B. Localization of Facial Features
Among the various facial features, eyes and mouth are
the most suitable features for recognition and estimation of
3D head pose. Most approaches to eye and face
localization are template based. However, our approach is
able to directly locate eyes, mouth, and face boundary
based on measurements derived from the color-space
components of an image.
[3] R.A. Redner, and H.F. Walker, “Mixture densities,
maximum likelihood and the EM algorithm”, SIAM
Review, Vol. 26, no. 2, pp. 195-239, 1984.
[4]C. Bishop, Neural Networks for Pattern Recognition,
Oxford University Press, 1995.
[5]www.theguardian.com/technology/2013/nov/11/tesc
o.
[6] Chellappa, Wilson, Sirohey, “Human and Machine
Recognition of Faces: A Survey”, Proceedings of the
IEEE, Vol. 83, No. 5, May 1995, pp. 705-740.
Fig.4 Example of Face Capture
[7]Judith A.blake(the Jackson laboratory U.S.A).
[8]Joseph R.Ecker Howard Hughes Medical Institute
and Salk Institute ,U.SA.
VI.
CONCLUSION
We have presented a face detection algorithm for color
images using a skin-tone color model and facial
features. Our method first corrects the color bias by a novel
lighting compensation technique that automatically
estimates the reference white pixels. We overcome the
difficulty of detecting the low-luma and high-luma skin
tones by applying a nonlinear transform to the YCbCr
color space. Our method detects skin regions over the
entire image, and then generates face candidates based
on the spatial arrangement of these skin patches. The
algorithm constructs eye/mouth/boundary maps for
verifying each face candidate. Detection results for several
photo collections have been presented. Our goal is to
design a system that detects faces and facial features,
allows users to edit detected faces, and uses the facial
features as indices for retrieval from image and video
databases.
REFRENCES
[1] M. Turk and A.P. Pentland, “Face recognition
using eigenfaces”, Proc. CVPR, 1991, pp 586-593.
[2] Jones, Rehg, “Statistical Color Models with
Application to Skin Detection”, Tech-Rep. CRL 98/11,
Compaq Cambridge Research Lab 1998.