a la physique de l'information - Lisa - Université d'Angers
a la physique de l'information - Lisa - Université d'Angers
a la physique de l'information - Lisa - Université d'Angers
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
noises Z i in (3) are chosen Gaussian. Other measures of simi<strong>la</strong>rity (like<br />
the peak-signal-to-noise ratio), images and PDF for Z i have been tested.<br />
Results were quantitatively and qualitatively simi<strong>la</strong>r to the ones<br />
presented below. In Fig. 1, the simi<strong>la</strong>rity between the restored and<br />
original image, assessed by the normalised cross-covariance of (5),<br />
surpasses the c<strong>la</strong>ssical Perona-Malik process for all values tested for M.<br />
This <strong>de</strong>monstrates the possibility of improving the performance of the<br />
Perona-Malik process by injecting a nonzero amount of the M noises Z i<br />
in (4). Moreover, in Fig. 1, one can also notice that the convergence<br />
speed of the diffusion process is increased with the presence of the M<br />
noises Z i. This acceleration of the convergence is another benefit<br />
shown in Fig. 1, which is also perceptible within the images themselves<br />
(Fig. 2). The addition of noise leads to a better contrast and a better<br />
preservation of the structure, which are characterised by small gradient<br />
(e.g. the buildings in the background are more visible and the texture of<br />
the grass is preserved by comparison with the c<strong>la</strong>ssical Perona-Malik<br />
process). In Fig. 3, the number of iterations n of the diffusion processes<br />
of (3) and (4) is fixed. The normalised cross-covariance of (5)<br />
culminates at a maximum for an optimal nonzero noise level of the<br />
M noises injected in (4). This clearly <strong>de</strong>monstrates the possibility of a<br />
constructive role of noise in the diffusion process of (3) and (4) for an<br />
image restoration task.<br />
Fig. 3 Normalised cross-covariance of (5) against RMS amplitu<strong>de</strong> sZ of<br />
noise purposely injected in (4)<br />
– – – our modified version of Perona-Malik process of (3) and (4) for various<br />
number M of in<strong>de</strong>pen<strong>de</strong>nt noises Zi ——— c<strong>la</strong>ssical Perona-Malik process of (2)<br />
Number of iterations n fixed for all lines to n ¼ 75<br />
# IEE 2006 27 January 2006<br />
Electronics Letters online no: 20060180<br />
doi: 10.1049/el:20060180<br />
A. Histace and D. Rousseau (Laboratoire d’Ingénierie <strong>de</strong>s Systèmes<br />
Automatisées (LISA CNRS FRE 2656), 62 avenue Notre Dame du Lac,<br />
49000 Angers, France)<br />
E-mail: aymeric.histace@univ-angers.fr<br />
References<br />
1 Boncelet, C.G.: ‘Image noise mo<strong>de</strong>ls’ in ‘Bovik, A. (Ed.): ‘Handbook of<br />
image and vi<strong>de</strong>o processing’ (Aca<strong>de</strong>mic Press, New York, 2000)<br />
2 Maheshwari, O., and Ebenezer, D.: ‘Simultaneous removal of positive and<br />
negative impulses in images by using adaptive length recursive weighted<br />
median filter’, WSEAS Trans. Commun., 2005,4, (12), pp. 1350–1355<br />
3 Perona, P., and Malik, J.: ‘Scale-space and edge <strong>de</strong>tection using anistropic<br />
diffusion’, IEEE Trans. Pattern Anal. Mach. Intell., 1990, 12, (7),<br />
pp. 629–639<br />
4 Catte, F., Coll, T., Lions, P., and Morel, J.: ‘Image selective smoothing<br />
and edge <strong>de</strong>tection by nonlinear diffusion’, SIAM J. Appl. Math., 1992,<br />
29, (1), pp. 182–193<br />
5 Harmer, G.P., Davis, B.R., and Abbott, D.: ‘A review of stochastic<br />
resonance: circuits and measurement’, IEEE Trans. Instrum. Meas.,<br />
2002, 51, pp. 299–309<br />
6 Komprobst, P., Deriche, R., and Aubert, G.: ‘Image sequence analysis via<br />
partial differential equations’, J. Math. Imaging Vis., 1999, 11, (1),<br />
pp. 5–26<br />
7 Chapeau-Blon<strong>de</strong>au, F., and Rousseau, D.: ‘Enhancement by noise in<br />
parallel arrays of sensors with power-<strong>la</strong>w characteristics’, Phys. Rev. E,<br />
2004, 70, pp. 060 101(R), 1–4<br />
ELECTRONICS LETTERS 30th March 2006 Vol. 42 No. 7<br />
148/197