Proceedings of the 8th ACM International Workshop on Multimedia Information Retrieval 2006
DOI: 10.1145/1178677.1178703
|View full text |Cite
|
Sign up to set email alerts
|

Image classification using cluster cooccurrence matrices of local relational features

Abstract: Image classification systems have received a recent boost from methods using local features generated over interest points, delivering higher robustness against partial occlusion and cluttered backgrounds. We propose in this paper to use relational features calculated over multiple directions and scales around these interest points. Furthermore, a very important design issue is the choice of similarity measure to compare the bags of local feature vectors generated by each image, for which we propose a novel ap… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
7
0
2

Year Published

2007
2007
2020
2020

Publication Types

Select...
4
3
1

Relationship

1
7

Authors

Journals

citations
Cited by 18 publications
(9 citation statements)
references
References 22 publications
0
7
0
2
Order By: Relevance
“…UFR: University of Freiburg, Computer Science Dep., Freiburg, Germany. The Pattern Recognition and Image Processing group from the University Freiburg, Germany, submitted four runs using relational features calculated around interest points which are later combined to form cluster cooccurrence matrices [21]. Three different classification methods were used.…”
Section: Participating Groups and Methodsmentioning
confidence: 99%
“…UFR: University of Freiburg, Computer Science Dep., Freiburg, Germany. The Pattern Recognition and Image Processing group from the University Freiburg, Germany, submitted four runs using relational features calculated around interest points which are later combined to form cluster cooccurrence matrices [21]. Three different classification methods were used.…”
Section: Participating Groups and Methodsmentioning
confidence: 99%
“…This descriptor was introduced in the context of texture classification by Haralick et al (1973). We extended the method in (Setia et al, 2006) by setting into relation the cluster memberships of pairs of features derived from local detections. A similar idea was explored in (Amores et al, 2005).…”
Section: Related Workmentioning
confidence: 99%
“…We discussed this method in detail in (Setia et al, 2006). The following entries of the co-occurrence matrix capture the statisti- Fðc; c 0 ; d; aÞ ¼ #fðs 1 ; s 2 ÞjðLðs 1 Þ ¼ cÞ^ðLðs 2 Þ ¼ c 0 Þ ðD d < jjs 1 À s 2 jj 2 6 D dþ1 Þ ðA a < ]ðs 1 ; s 2 Þ 6 A aþ1 Þg; ð10Þ…”
Section: Orientation Variant Accumulatormentioning
confidence: 99%
“…On the other hand, images are usually taken under very standardised conditions and from well-defined view angles, so invariances are often different than in stock photography where invariance to particular lighting conditions is of interest. Still, some invariances and particularly the use of local interest points can prove to be successful [15]. Problems with the handling of medical images (not only for retrieval) occur at many levels: DICOM CT (Computed Tomography) or MR (Magnet Resonance Tomography) images are 12-14 Bit grey scale and thus more than what many image processing systems cope with and also more than can be shown screen.…”
Section: State Of the Artmentioning
confidence: 99%