Directly invertible nonlinear divisive normalization pyramid for image representation

Roberto Valerio, Eero Simoncelli, Rafael Navarro

Research output: Contribution to journalArticle

Abstract

We present a multiscale nonlinear image representation that permits an efficient coding of natural images. The input image is first decomposed into a set of subbands at multiple scales and orientations using near-orthogonal symmetric quadrature mirror filters. This is followed by a nonlinear "divisive normalization" stage, in which each linear coefficient is divided by a value computed from a small set of neighboring coefficients in space, orientation and scale. This neighborhood is chosen to allow this nonlinear operation to be efficiently inverted. The parameters of the normalization operation are optimized in order to maximize the independence of the normalized responses for natural images. We demonstrate the near-independence of these nonlinear responses, and suggest a number of applications for which this representation should be well suited.

Original languageEnglish (US)
Pages (from-to)331-340
Number of pages10
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2849
StatePublished - 2003

Fingerprint

Image Representation
Pyramid
Digital filters
Invertible
Normalization
Multiple Scales
Nonlinear Response
Coefficient
Quadrature
Mirror
Coding
Maximise
Filter
Demonstrate
Independence

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

@article{5081e9e611604a6c9620aa972eb0e7a2,
title = "Directly invertible nonlinear divisive normalization pyramid for image representation",
abstract = "We present a multiscale nonlinear image representation that permits an efficient coding of natural images. The input image is first decomposed into a set of subbands at multiple scales and orientations using near-orthogonal symmetric quadrature mirror filters. This is followed by a nonlinear {"}divisive normalization{"} stage, in which each linear coefficient is divided by a value computed from a small set of neighboring coefficients in space, orientation and scale. This neighborhood is chosen to allow this nonlinear operation to be efficiently inverted. The parameters of the normalization operation are optimized in order to maximize the independence of the normalized responses for natural images. We demonstrate the near-independence of these nonlinear responses, and suggest a number of applications for which this representation should be well suited.",
author = "Roberto Valerio and Eero Simoncelli and Rafael Navarro",
year = "2003",
language = "English (US)",
volume = "2849",
pages = "331--340",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - Directly invertible nonlinear divisive normalization pyramid for image representation

AU - Valerio, Roberto

AU - Simoncelli, Eero

AU - Navarro, Rafael

PY - 2003

Y1 - 2003

N2 - We present a multiscale nonlinear image representation that permits an efficient coding of natural images. The input image is first decomposed into a set of subbands at multiple scales and orientations using near-orthogonal symmetric quadrature mirror filters. This is followed by a nonlinear "divisive normalization" stage, in which each linear coefficient is divided by a value computed from a small set of neighboring coefficients in space, orientation and scale. This neighborhood is chosen to allow this nonlinear operation to be efficiently inverted. The parameters of the normalization operation are optimized in order to maximize the independence of the normalized responses for natural images. We demonstrate the near-independence of these nonlinear responses, and suggest a number of applications for which this representation should be well suited.

AB - We present a multiscale nonlinear image representation that permits an efficient coding of natural images. The input image is first decomposed into a set of subbands at multiple scales and orientations using near-orthogonal symmetric quadrature mirror filters. This is followed by a nonlinear "divisive normalization" stage, in which each linear coefficient is divided by a value computed from a small set of neighboring coefficients in space, orientation and scale. This neighborhood is chosen to allow this nonlinear operation to be efficiently inverted. The parameters of the normalization operation are optimized in order to maximize the independence of the normalized responses for natural images. We demonstrate the near-independence of these nonlinear responses, and suggest a number of applications for which this representation should be well suited.

UR - http://www.scopus.com/inward/record.url?scp=24644510240&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=24644510240&partnerID=8YFLogxK

M3 - Article

VL - 2849

SP - 331

EP - 340

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

ER -