Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group

Octave-wide channels and their relation to natural images

Not Accessible

Your library or personal account may give you access

Abstract

One of the parameters that describes the spatial properties of cells in the mammalian visual cortex is the spatial-frequency bandwidth. Although there is some variation in the range of tuning (0.5–2.5 octaves), the physiological evidence supports the notion that the bandwidths of cortical cells increase as the peak sensitivity increases in spatial frequency; the bandwidths are roughly constant (in octaves) on a log-frequency axis. This study looks into the possible reasons for this type of representation in the mammalian system. One suggestion is that octave-wide bandwidths allow for an equal distribution of activity throughout the sensors of the code. This can be shown to be true for channels with equal peak sensitivities and images with amplitude spectra of 1/f-However, there are a number of ways to produce this equal distribution other than by octave-wide codes. In this study, a second possibility was explored. By using a Gabor-like code and a variety of calibrated natural scenes, statistical properties were described in terms of a space versus frequency representation. In such a representation, features in natural images (local regions where the phases are correlated) become spatially smaller and broader in spatial frequency with increasing frequency. This is not a necessary property of scenes in general, and some images will be described that do not show this property. Such images may be more effectively represented by codes that do not show octave-wide bandwidths.

© 1990 Optical Society of America

PDF Article
More Like This
Parameters that influence the distribution of activity in visual codes

Nuala Brady and David J. Field
TuY14 OSA Annual Meeting (FIO) 1990

Texture coding and analysis by Gabor functions

Antonio Tabernero and Rafael Navarro
MT3 OSA Annual Meeting (FIO) 1990

Toward a theory of face recognition

Matthew A. Turk and Alex P. Pentland
MEE3 OSA Annual Meeting (FIO) 1990

Select as filters


Select Topics Cancel
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.