In developing computer vision applications namely Human
Computer interaction system and Hand Gesture based system,
skin color detection plays a crucial role. This is the first step in
developing these systems. But this step becomes challenging
due to the factors like variable lightening conditions, complex
background, skin colored textured objects etc. So skin color
detection algorithm must be designing with the capability to
work well under these circumstances. Broadly classifying skin
color detection algorithm comprises of two aspects namely:
skin color representation and skin color model. Further skin
color can be represented as basic color space or perceptual
color space or orthogonal color space or perceptually uniform
color space. Similarly skin color models are classified into
three broad categories namely Pixel-based, Edge based and
region based skin segmentation. Pixel based models include
Gaussian classifier, histogram based, Bayesian classifier
explicitly defined region. Similarly, edge based models are
watershed regions, Laplacian of Gaussian, canny edge
detector. Finally region based models are region splitting,
region merging and region growing techniques [6].
The proposed work is based on the explicitly defined region
model based on pixel based skin segmentation and color space
used here is YCbCr. Real time human skin color detection
suffers from mainly two problems: complex background and
lightning factors. In order to remove complex background a
Skin map is created to protect skin color pixels from
background, known as foreground representation and then
apply background subtraction to obtain the segmented skin
colored image as a desired output. Fig 5 shows the proposed
algorithm for Human skin color detection using skin map is
given in figure 7.