Interactive access to large image collections using similarity-based visualization

Nguyen, G.P.; Worring, Marcel

doi:10.1016/j.jvlc.2006.09.002

Cited by 110 publications

(68 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Mapping-based techniques employ dimensionality reduction techniques to map high-dimensional image feature vectors to a low-dimensional space for visualisation. Typical examples examples use principal component analysis (PCA) [10,11], multi-dimensional scaling (MDS) [12], or non-linear embedding techniques [13] to define a visualisation space onto which to place images. Clustering-based visualisations group visually similar images together, often in a hierarchical manner.…”

Section: Image Database Browsingmentioning

confidence: 99%

Intuitive Image Database Browsing to Support Image Annotation

Schaefer

Stuttard²

2012

Automatika

View full text Add to dashboard Cite

Original scientific paperImage annotation is typically performed manually since automatic image annotation approaches have not matured yet to be used in practice. Consequently, image annotation is a labour intensive and time consuming task. In this paper, we show how an image browsing system can be employed to support efficient and effective (manual) annotation of image databases. In contrast to other approaches, which typically present images in a linear fashion, we employ a visualisation where images are arranged by mutual visual similarity. Since in this arrangement similar images are close to each other, they can easily be selected and annotated together. Organisation on a grid layout prevents image overlap and thus contributes to a clear presentation. Large image databases are handled through a hierarchical data structure where each image in the visualisation can correspond to a cluster of images that can be expanded by the user. Experimental results indicate that annotation can be performed faster on our proposed system. Key words:Visual information processing, Image databases, Image browsing, Image annotation, Hue sphere.Intuitivno pretraživanje baze slike kao potpora označavanju slika. Označavanje slika obično se obavlja ručno jer automatski pristupi još nisu dovoljno kvalitetni kako bi se koristili u praksi. Zbog toga je označavanje slika u bazi vremenski zahtjevno. U ovom radu pokazatćemo kako se sustav za pregled slika u bazi može koristiti kao učinkovita potpora ručnom označavanju slika. Za razliku od drugih pristupa, koji prikazuju slike u linearnom poretku, korištena je vizualizacija u kojoj su slike složene po medusobnoj sličnosti. Budući da su na taj način slične slike medusobno blizu jedna drugoj, lako ih je selektirati i zajednički označiti. Slike su organizirane u mrežni prikaz radi sprječavanja preklapanja i jasnije prezentacije. Velike baze podataka organizirane su u hijerarhijsku strukturu gdje svaka slika u pojedinoj vizualizaciji može pripadati skupu slikačiji prikaz korisnik po želji može proširivati. Rezultati provedenih eksperimenata pokazuju da se označavanje slika pomoću predloženog sustava može obavljati brže nego na uobičajeni način.Ključne riječi: obrada vizualnih informacija, baze slika, pretraživanje slika, označavanje slika

show abstract

Section: Image Database Browsingmentioning

confidence: 99%

Intuitive Image Database Browsing to Support Image Annotation

Schaefer

Stuttard²

2012

Automatika

View full text Add to dashboard Cite

show abstract

“…We denote the set of images displayed in iteration as . This set should be carefully chosen as it affects the resulting , most importantly should give an adequate overview of the whole collection [16]. First, the collection is divided into a set of clusters using a clustering algorithm.…”

Section: A Creation Of the Dissimilarity Spacementioning

confidence: 99%

“…A projection from the high dimensional space, being it feature space or dissimilarity space, to two dimensions is needed. Similarity-based visualization, [14], [21], [16], [20] works on dissimilarity space directly: (13) With a matrix containing dissimilarities, which can be either the original dissimilarity space or the prototype-based dissimilarity space. In similarity-based visualization, the position of images in is chosen such that distances reflect, as faithfully as possible, the dissimilarities in dissimilarity space .…”

Section: B the Manipulation Spacementioning

confidence: 99%

“…In [16], we have compared five different projection techniques namely Isomap (Isometric mapping), Stochastic Neighbor Embedding (SNE), Local Linear Embedding (LLE), Multi-Dimensional scaling (MDS), and a new method ISOSNE which replaces the MDS step used in Isomap by SNE. The projection method giving the best performance in terms of preserving original distances is ISOSNE, so this is used as in this paper.…”

Section: B the Manipulation Spacementioning

confidence: 99%

“…Relations are explicit in [3] and [11], where the existence of discrete relations between images are presented as a graph. The method in [16] does not use discrete relations, but visualizes the similarity among the images. More methods along this line can be found in the reference.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Interactive Search by Direct Manipulation of Dissimilarity Space

Nguyen

Worring

Smeulders

2007

IEEE Trans. Multimedia

Self Cite

View full text Add to dashboard Cite

Abstract-In this paper, we argue to learn dissimilarity for interactive search in content based image retrieval. In literature, dissimilarity is often learned via the feature space by feature selection, feature weighting or by adjusting the parameters of a function of the features. Other than existing techniques, we use feedback to adjust the dissimilarity space independent of feature space. This has the great advantage that it manipulates dissimilarity directly. To create a dissimilarity space, we use the method proposed by Pekalska and Duin, selecting a set of images called prototypes and computing distances to those prototypes for all images in the collection. After the user gives feedback, we apply active learning with a one-class support vector machine to decide the movement of images such that relevant images stay close together while irrelevant ones are pushed away (the work of Guo et al.). The dissimilarity space is then adjusted accordingly. Results on a Corel dataset of 10000 images and a TrecVid collection of 43907 keyframes show that our proposed approach is not only intuitive, it also significantly improves the retrieval performance.Index Terms-Active learning, dissimilarity learning, interactive image search, visualization.

show abstract

Multimedia Analytics for Image Collection Forensics

Worring

2015

Handbook of Digital Forensics of Multimedia Data and Devices

View full text Add to dashboard Cite

Interactive access to large image collections using similarity-based visualization

Cited by 110 publications

References 20 publications

Intuitive Image Database Browsing to Support Image Annotation

Intuitive Image Database Browsing to Support Image Annotation

Interactive Search by Direct Manipulation of Dissimilarity Space

Multimedia Analytics for Image Collection Forensics

Contact Info

Product

Resources

About