08-09-2014, 02:35 PM
An image processing analysis of skin textures
An image processing.pdf (Size: 735.69 KB / Downloads: 97)
Abstract
Colour and coarseness of skin are visually different. When image processing is involved in
the skin analysis, it is important to quantitatively evaluate such differences using texture
features. In this paper, we discuss a texture analysis and measurements based on a statistical
approach to the pattern recognition. Grain size and anisotropy are evaluated with proper
diagrams. The possibility to determine the presence of pattern defects is also discussed.
Introduction
A quantitative characterisation of human skin textures is one of the tasks recently approached
by image processing. This problem of texture analysis is twofold interesting. Besides the
computational modelling of skin for realistic rendering in computer graphics [1], we must
consider the possibility to apply the texture analysis to computer-assisted diagnosis in
dermatology [2,3].
Image analysis
determined:
( ) ∫ ∫ =
x
l y
l
x y
o b y,x dxdy
ll
M
0 0
1
(1)
where x y
l l, are the x- and y- rectangular range of the image frame. More generally, the k-rank
statistical moments of the image are defined in the following way:
[ ] ( ) ∫ ∫ = −
x
l y
l
k
o
x y
k
b y,x M dxdy
l l
M
0 0
1
(2)
With this kind of characterisation we are then able to define the average values of the
moments for the whole image frame. The distribution of pixel tones is then given according to
these moments. The tone dispersion turns out to be evaluate by moment with k=2.
All integrals can be calculated on the whole image or on a window
Analysis of skin textures and discussion
The case of snake skin gives diagrams immediately recalling the properties of the unit cells in
crystal lattices. This is because the texture is quite geometric. The features of human skin are
of course different, but as we observe from its use in microscopy investigation of liquid
crystals [13,14], it is in the case of almost homogenous images, where Fourier analysis is
scarcely active, that coherence lengths are useful.
We analysed with the coherence length ( L i,o
) diagrams images of human-like textures and the
result in shown in Fig.3, in the middle of the figure. In the lower part of the figure, it is shown
a leather texture and the corresponding analysis. The inner and outer curves have thresholds
as those used to obtain diagrams of Fig.2. The shape of the two diagrams does not
substantially change. The area is changing: this is due to the fact that, to be fulfilled, a lower
threshold requires a wider area.
On the right part of the figure, we see the detection of defects: the points marked in red are
considered as "defects", whereas the pixels in green are the normal ones. It is not a
segmentation procedure at the origin of the red-green maps, but a criterion involving the local
behaviour of the coherence lengths l ( y,x ) i,o
. In the case that we are discussing