Self-adaptive algorithm for segmenting skin regions

EURASIP Journal on Advances in Signal Processing

Table 1 The symbols used in the paper

Symbol	Description
General symbols
	Input color image
$C_{s}$	Skin class
$C_{ns}$	Non-skin class
P	Probability
$P_{S}$	Skin probability map
	Histogram
Σ	Covariance matrix
	A set of blobs
$N_{n \times n} (x)$	A set of pixels in the n×n kernel around thepixel x
δ _fp	False-positive rate
$δ_{min}$	Minimal error
Symbols related with
the DT-based spatial analysis
(including the DSPF space)
Γ	Total path cost
γ	Overall local cost
γ _p	Destination-probability local cost component
γ _Δ	Local cost component related with thedifference in the propagation domain
$T_{α}^{P}$	High-probability seed extractionthreshold
$T_{β}^{P}$	Lower-bound propagation threshold
$T_{0}^{P}$	Costless propagation threshold
$T_{Γ}$	Total path cost threshold
	Distance in the DSPF space
ν	Feature vector in the DSPF space
r	Reference pixel
P_r	Reference skin probability (determinedin the neighborhood of r)
Symbols related
with the self-adaptive
seeds method
$S^{0}$	Initial skin seeds
$S^{E}$	Expanded skin seeds
$S^{A}$	Adapted (final) skin seeds
t _seed	Dynamic initial binarization threshold
$R_{seed}$	Ratio of pixels used to determine t_seed
$T_{A}^{P}$	Binarization threshold used to extract $S^{A}$ from the local skinprobability map
$T_{seed}^{P}$	Minimum acceptable value of t_seed
$T_{r}^{P}$	Threshold for the reference skin probability(in the dilated $P_{S}$ )
$γ_{Δ}^{E}$	Local difference costs (γ_Δ) used for buildingthe expanded
$γ_{Δ}^{F}$	Local difference costs (γ_Δ) used for final‘skinness’ propagation