Publications
2018
1.
Olivera, Óscar García-Olalla; Alegre, Enrique; Fernández-Robles, Laura; Fidalgo, Eduardo; Saikia, Surajit
Textile retrieval based on image content from cdc and webcam cameras in indoor environments Artículo de revista
En: Sensors, vol. 18, no 5, pp. 1329, 2018, (Publisher: MDPI).
Resumen | Enlaces | BibTeX | Etiquetas: Content-based Image Retrieval, Textile Localization, Textile Retrieval, texture description, Texture Retrieval, Visual Sensors
@article{garcia-olalla_olivera_textile_2018,
title = {Textile retrieval based on image content from cdc and webcam cameras in indoor environments},
author = {Óscar García-Olalla Olivera and Enrique Alegre and Laura Fernández-Robles and Eduardo Fidalgo and Surajit Saikia},
url = {https://www.mdpi.com/1424-8220/18/5/1329},
year = {2018},
date = {2018-01-01},
urldate = {2018-01-01},
journal = {Sensors},
volume = {18},
number = {5},
pages = {1329},
abstract = {This paper presents a method for textile-based image retrieval, useful for law enforcement to identify evidence by matching textiles in real-world scenes. The approach uses MSER on high-pass filtered RGB, HSV, and Hue channels to extract textile regions. HOG and HCLOSIB are combined for feature description and correlation distance to match the query textile patch with candidate regions. A new dataset, TextilTube, consisting of 1913 labeled textile regions in 67 classes, is introduced. Experimental results show 84.94% success in the top 40 matches and 37.44% precision for the first match, outperforming current deep learning methods.},
note = {Publisher: MDPI},
keywords = {Content-based Image Retrieval, Textile Localization, Textile Retrieval, texture description, Texture Retrieval, Visual Sensors},
pubstate = {published},
tppubtype = {article}
}
This paper presents a method for textile-based image retrieval, useful for law enforcement to identify evidence by matching textiles in real-world scenes. The approach uses MSER on high-pass filtered RGB, HSV, and Hue channels to extract textile regions. HOG and HCLOSIB are combined for feature description and correlation distance to match the query textile patch with candidate regions. A new dataset, TextilTube, consisting of 1913 labeled textile regions in 67 classes, is introduced. Experimental results show 84.94% success in the top 40 matches and 37.44% precision for the first match, outperforming current deep learning methods.