Noise processes are often modelled as stochastic processes. We have used a multivariate method based on the application of Principal Component Analysis (PCA) in order to classify different spatial-temporal structures taken as noise. When the structures have a correlation in time, a parameter distinguishing between fast and slow dynamics appears naturally. We have found this parameter in previous contributions with a different meaning depending on the context. Especially interesting is the application to the characterization of 1/f noise. In this paper we have extended the method in order to apply it to different kind of systems exhibiting, for example, self-organizing properties or brownian motion. One goal is trying to define a criterion to distinguish between fast and slow dynamics parameters. Finally, a statistical analysis is made in order to find the conditions for the application of the method to a wide range of different systems.
[1]
Javier Alda,et al.
Principal-component characterization of noise for infrared images.
,
2002,
Applied optics.
[2]
Edward R. Dougherty,et al.
Random processes for image and signal processing
,
1998,
SPIE / IEEE series on imaging science and engineering.
[3]
J. Walkup,et al.
Statistical optics
,
1986,
IEEE Journal of Quantum Electronics.
[4]
J M Mooney.
Effect of spatial noise on the minimum resolvable temperature of a staring sensor.
,
1991,
Applied optics.
[5]
D. F. Morrison,et al.
Multivariate Statistical Methods
,
1968
.
[6]
Javier Alda,et al.
Operational parametrization of the 1/f noise of a sequence of frames by means of the principal component analysis in focal plane arrays
,
2003
.