Publications
2013
1.
García-Ordás, María Teresa; Alegre, Enrique; Olivera, Óscar García-Olalla; García-Ordás, Diego
Evaluation of different metrics for shape based image retrieval using a new contour points descriptor Artículo de revista
En: Similarity Search and Applications: 6th International Conference, SISAP 2013, A Coruña, Spain, October 2-4, 2013, Proceedings 6, pp. 141–150, 2013, (Publisher: Springer Berlin Heidelberg).
Resumen | Enlaces | BibTeX | Etiquetas: CPDH36R descriptor, kNN, Shape Retrieval
@article{garcia-ordas_evaluation_2013,
title = {Evaluation of different metrics for shape based image retrieval using a new contour points descriptor},
author = {María Teresa García-Ordás and Enrique Alegre and Óscar García-Olalla Olivera and Diego García-Ordás},
url = {https://link.springer.com/chapter/10.1007/978-3-642-41062-8_14},
year = {2013},
date = {2013-01-01},
journal = {Similarity Search and Applications: 6th International Conference, SISAP 2013, A Coruña, Spain, October 2-4, 2013, Proceedings 6},
pages = {141–150},
abstract = {This paper evaluates an image shape retrieval method using various distance metrics (Euclidean, Intersect, Hamming, and Cityblock) and k-nearest neighbors (kNN) classifiers (original kNN, mean distance kNN, and weighted kNN). A new shape descriptor, CPDH36R, improves upon the original CPDH descriptor, achieving higher success rates across Kimia99, Kimia25, MPEG7, and MPEG2 datasets. The Cityblock distance with the original kNN classifier yields the best performance, notably improving accuracy in MPEG2. Additionally, the proposed method significantly reduces computational cost compared to the original Earth Mover Distance classifier.},
note = {Publisher: Springer Berlin Heidelberg},
keywords = {CPDH36R descriptor, kNN, Shape Retrieval},
pubstate = {published},
tppubtype = {article}
}
This paper evaluates an image shape retrieval method using various distance metrics (Euclidean, Intersect, Hamming, and Cityblock) and k-nearest neighbors (kNN) classifiers (original kNN, mean distance kNN, and weighted kNN). A new shape descriptor, CPDH36R, improves upon the original CPDH descriptor, achieving higher success rates across Kimia99, Kimia25, MPEG7, and MPEG2 datasets. The Cityblock distance with the original kNN classifier yields the best performance, notably improving accuracy in MPEG2. Additionally, the proposed method significantly reduces computational cost compared to the original Earth Mover Distance classifier.