particular field based on computer-human interaction is face detection. Face
recognition is a biometrics pattern authentication technique which is used in
a wide variety of computer vision applications. It plays a major role in video
surveillance, security, digital video processing, content-based retrieval, etc.
The objective of face detection is to detect any faces in an arbitrary image
and return the image location and extent of each face. There are many
factors that make the real-time face detection a challenging task. One of the
main problems is the time and accuracy which determine the performance of
face recognition system in real-time environments. Additionally, variations in
Image
preporcessin
g
Applying
Principial
Component
Analysis
Adaptive
Histoqram
Equalization
Threshold
Level
Determinatio
n
Converting
the
Subtracted
Image to the
Binary Image
Colorizing
the final
result
THE 3
rd
INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS
dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
180
scale, location, view point, illumination, occlusions can affect the overall
accuracy of system.
There are many methods and algorithms available for face detection,
such as Viola-Jones, SMQT features &SNOW Classifier, Neural Network-
Based Face Detection and Support Vector Machine-Based face detection.
Although all these algorithms manage to detect faces from an image, their
precision and recall values are different. This paper focuses on Viola and
Jones face detection algorithm which has the high image processing speed
and detection rates.
Viola-Jones algorithm is implemented in OpenCV and combines four
main concepts:
1.
Haar like features
Every human face has certain common properties. For example, eye
region is darker than nose bridge region or upper cheek region is brighter
than eye region, etc. Haar like features detect the difference in the black and
light portion of the image and this computation forms a single rectangle
around the face. There are some commonly used Haar features, like two
rectangle feature, three rectangle feature, etc.
2.
Integral image
Integral image is an intermediate representation for the image and
allows to compute the sum of values in a rectangle subset of a pixel grid in a
very efficient and quick way. The integral image at location x, y contains the
sum of the pixels above and to the left of x, y, inclusive:
where i(x, y) is the pixel value of the original image and ii(x’,y’) is the
corresponding image integral value.
Figure 1.
Illustration of the integral image and 6 types of Haar-like rectangle features
|