Image representation and reconstruction in the SE(2) gaussian wavelets model of primary visual cortex
Abstract. The purpose of this talk is to present some new insights on a classical model of visual neurons in brain's primary visual cortex V1 from the point of view of harmonic analysis.
Most of the linear behaviour of a single V1 neuron can be described as a wavelet coefficient of the visual stimulus, associated with the quasiregular representation of the SE(2) group on a gaussian mother wavelet. However, the set of available coefficients (as observed with neurophysiological measurements on V1) has a puzzling distribution, which does not immediately resemble known sampling sets. We want to present - with analytical and numerical arguments - an approach that can allow us to obtain frame conditions, and quantify the accuracy of the representation, for a relevant function space of images. We will also show an elementary reconstruction algorithm that does not require the explicit computation of a dual frame.
This series of seminars is addressed to an audience interested in Harmonic Analysis in the broadest possible sense. The seminars will not necessarily concern the latest research results; the speaker may also give a talk about open problems or a survey colloquium.
The conferences take place generally every two weeks on Wednesday at 5:30 p. m. (Rome time).
To subscribe to our mailing list please write to seminarivaa@gmail.com.
Organizers:
Tommaso Bruno (Università di Genova)
Valentina Casarino (Università degli Studi di Padova)
Bianca Gariboldi (Università degli Studi di Bergamo)
Alessio Martini (Politecnico di Torino)