Automatic visual features weights obtention for Content-Based Image Retrieval Systems

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

In Content-Based Image Retrieval (CBIR) Systems it is necessary to combine more than one visual descriptor in order to improve the retrieval performance. The most common descriptors are Color-Based, Shape-Based and Texture-Based descriptors. When more than one visual descriptor is linearly combined, some adequate weight must be assigned to each visual feature. The most common manner is setting the same weight value for each visual feature. The sum of these values must be equal to one. However, this process does not guarantee the optimum performance of the CBIR system. In order to guarantee the best performance, it is necessary to do several experimentations to find the optimum weight values combination. This is time consuming process and ambiguous, due to the weights values depends on the nature of the databases. In this paper we proposed a scheme which computes automatically the best weight combination and guarantees the optimum performance of the CBIR system.

Original languageEnglish
Title of host publication2015 12th International Conference on Electrical Engineering, Computing Science and Automatic Control, CCE 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781467378390
DOIs
StatePublished - 14 Dec 2015
Event12th International Conference on Electrical Engineering, Computing Science and Automatic Control, CCE 2015 - Mexico City, Mexico
Duration: 26 Oct 201530 Oct 2015

Publication series

Name2015 12th International Conference on Electrical Engineering, Computing Science and Automatic Control, CCE 2015

Conference

Conference12th International Conference on Electrical Engineering, Computing Science and Automatic Control, CCE 2015
Country/TerritoryMexico
CityMexico City
Period26/10/1530/10/15

Keywords

  • CBIR
  • Visual Descriptors
  • Weighted Linear Combination
  • Weighted Visual Features

Fingerprint

Dive into the research topics of 'Automatic visual features weights obtention for Content-Based Image Retrieval Systems'. Together they form a unique fingerprint.

Cite this