Vis enkel innførsel

dc.contributor.advisorKarlsen, Randi
dc.contributor.authorEvertsen, Martin Hætta
dc.date.accessioned2010-08-27T07:25:47Z
dc.date.available2010-08-27T07:25:47Z
dc.date.issued2010-06
dc.description.abstractPeople love to take images, but are not so willing to annotate the images af-terwards with relevant tags. Manually tagging images is both subjective (dependent on annotator) and time consuming. It would be nice if the tag-ging process could be done automatically. A requirement for effective searching and retrieval of images in rapid growing online image databases is that each image has accurate and useful annotation. This thesis shows that automatic tagging of images with relevant tags is possible by using a combination of the capture location, the date/time when the image was captured and an image category. The use of image categories (together with location and date/time) ensures that many relevant tags are returned and restrict the occurrence of noisy tags to a very low level despite using a noisy image database (Flickr). Other methods used for further re-stricting noise are to restrict usage of more than one image from same user (as basis for tagging the query image) and a dynamic approach for using many images when possible, and fewer images when not many relevant im-ages are found. The designed system is able to tag an image as long as there are a sufficient number of geo-referenced and already tagged images that is relevant for the query image available on Flickr. The query image must also have been geo-referenced and it is assumed that the user provides an image category. Im-ages are processed based on which category the images belongs to, i.e. an image is processed with the best method to handle images belonging to that specific category. In short, this means that images of objects or places are processed differently than images from events. The evaluation of the system indicates that usage of image categories is very helpful when tagging images. The system finds more relevant tags and fewer noisy tags than baseline systems using only location. It also performs good compared to a system using both location and content-based image analysis.en
dc.format.extent2347702 bytes
dc.format.mimetypeapplication/pdf
dc.identifier.urihttps://hdl.handle.net/10037/2633
dc.identifier.urnURN:NBN:no-uit_munin_2379
dc.language.isoengen
dc.publisherUniversitetet i Tromsøen
dc.publisherUniversity of Tromsøen
dc.rights.accessRightsopenAccess
dc.rights.holderCopyright 2010 The Author(s)
dc.rights.urihttps://creativecommons.org/licenses/by-nc-sa/3.0en_US
dc.rightsAttribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)en_US
dc.subject.courseIDINF-3981nor
dc.subjectVDP::Matematikk og naturvitenskap: 400::Informasjons- og kommunikasjonsvitenskap: 420::Simulering, visualisering, signalbehandling, bildeanalyse: 429en
dc.subjectVDP::Teknologi: 500::Informasjons- og kommunikasjonsteknologi: 550::Annen informasjonsteknologi: 559en
dc.subjectComputer Graphicsen
dc.subjectInformatikken
dc.titleAutomatic Image Tagging based on Context Informationen
dc.typeMaster thesisen
dc.typeMastergradsoppgaveen


Tilhørende fil(er)

Thumbnail
Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel

Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)
Med mindre det står noe annet, er denne innførselens lisens beskrevet som Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)