APPLICATION 1: FAST MULTI-FACE DETECTION IN COLOR IMAGES

USING FUZZY LOGIC:

SUMMARY:

Human face detection plays an important role in applications such as video surveillance, human
computer interface, face recognition, and face image database management, etc. The algorithm is
mainly based on skin color, sizes, shapes of faces and facial features are fuzzy factors
for verifying face candidates.

Many reported methods assume that the faces in an image or an image sequence have been
identified and localized. To build fully automated systems that analyze the information contained
in face images, robust and efficient face detection algorithms are required. Given an image, the
goal of face detection is to identify all image regions which contain a face regardless of its three-
dimensional position, orientation, and lighting conditions. Such a problem is challenging because
faces are not rigid and have a high degree of variability in size, shape, color, and texture.

The following paper proposes a face detection algorithm that is able to handle a wide range of
variations in still color images. Based on the compact distribution of human skin colors in
YCbCr color space, the proposed method detects skin regions over the entire image, and then
generates face candidates based on these skin regions' borders. The algorithm uses a fuzzy
reasoning engine with fuzzy factors such as the shape, the size of a face, and the presence of eyes
and lips on a face to verify each face candidate. In this process, Fast Marching method is used to
separate connected face regions. At the end of the algorithm, to conserve the information on
faces, Graham Scan method is used to construct convex hulls for true face borders. As the
results, relatively accurate borders of the true faces are obtained.

In the beginning of the algorithm, color segmentation is used to classify skin color pixels and
non-skin color pixels in the input image. A human face has a wide range of variations in a color
image. Therefore, to get high detection rate and low false positives, we need a flexible face
detection algorithm. The fuzzy logic approach is appropriate. The verification process for face
candidates is based on the structure of a fuzzy controller. The shape, the size of a face, and the
presence of eyes, lips or noses become the fuzzy factors of the fuzzy reasoning. Based on
different characteristics, a statistical survey is conducted to build fuzzy membership functions of
different fuzzy factors. The survey also gives some special cases such as a face connected with
background or faces connected via the skin color, and a skin region containing no hole (inside
border).
A true face may not contain any inside border since its features have the same skin color. But
these features usually have darker luminance and have horizontal shapes. A horizontal edge filter
has been used to find these objects. After that, the borders of these features again and consider its
fuzzy factors such as their sizes and positions relative to the outside border containing them.

Experimental results demonstrate successful face detection over a wide range of facial variations
in color, position, scale, orientation, 3D pose, and expression in images from several photo
collections.