Contributions to ICA of natural images

Author

Martin-Clemente, Ruben ; Hornillo-Mellado, Susana

Author_Institution

Dipt. de Teor. de la Senal y Comun., Univ. of Seville, Seville, Spain

fYear

2006

fDate

4-8 Sept. 2006

Firstpage

1

Lastpage

5

Abstract

In this paper we analyze the results provided by the popular algorithm FastICA when it is applied to natural images, using the kurtosis as non-linearity. In this case show that the so-called ICA filters can be expressed in terms of the eigenvectors associated to the smallest eigenvalues of the data correlation matrix, meaning that these filters are all high-pass. From this property emerges the sparse distribution of the independent components. On the other hand, the use of the kurtosis as contrast function causes the appearance of “spikes” in the independent components that make that the ICA bases are very similar to patches of the images analyzed. Some experiments are included to illustrate the results.

Keywords

correlation theory; eigenvalues and eigenfunctions; high-pass filters; image filtering; independent component analysis; natural scenes; statistical distributions; FastiCA; ICA filter; contrast function; data correlation matrix; eigenvalues; eigenvectors; high-pass filters; independent component analysis; kurtosis; natural image; sparse distribution; spikes; Abstracts; Propulsion;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing Conference, 2006 14th European

Conference_Location

Florence

ISSN

2219-5491

Type

conf

Filename

7071336