Image enhancement by path partitioning

Mario Lucertini, Yehoshua Perl, Bruno Simeone

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Scopus citations

Abstract

Image segmentation can be a useful tool in facing image degradation. In image segmentation the input is a set of pixels with given grey levels and the output is a partition of the set of pixels into connected regions ("classes"), so that a given set of requirements on the single classes and on adjacent classes is satisfied (i.e. pixels belonging to the same class must have approximately the same grey levels or the same textures and pixels belonging to adjacent classes must have significantly different grey levels or different textures). Once segmentation has been performed, the same grey level is associated with each pixel of the same class. The grey level can either be related to the original grey levels of the class, or can be given by a new grey scale on the ground of contrast optimization criteria. The segmentation technique proposed in this presentation is a method for finding the most homogeneous classes and the best possible contrast in a row by row image processing. In partitioning each row of the image, we have two aims: the partition must be as good as possible in its own right, and it must be as compatible as possible with the partitions of the other rows. If we take into account the two aims simultaneously, then the solution procedure becomes complex. To simplify and speed-up the procedure, we can partition each row independently, and then we can apply region merging techniques to the resulting set of row partitions. In the presentation the problem is formulated as a path partitioning one and a simple O(n p) row-partitioning algorithm based on a shortest path formulation of the problem is given.

Original languageEnglish (US)
Title of host publicationRecent Issues in Pattern Analysis and Recognition
EditorsVirginio Cantoni, Stefano Levialdi, Reiner Creutzburg, Gottfried Wolf
PublisherSpringer Verlag
Pages12-22
Number of pages11
ISBN (Print)9783540518150
DOIs
StatePublished - Jan 1 1989
EventBiannual International Conference on Pattern Recognition, 1988 - Rome, Italy
Duration: Nov 14 1988Nov 17 1988

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume399 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

OtherBiannual International Conference on Pattern Recognition, 1988
CountryItaly
CityRome
Period11/14/8811/17/88

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Image enhancement by path partitioning'. Together they form a unique fingerprint.

Cite this