Etiquetado automático de imágenes digitales mediante un algoritmo de ensamble semi-supervisado

Content-based image retrieval is a technique that uses the visual content of images such as color, form or texture to search images from large image collections. One way to carry out content-based image retrieval consists of image annotation, that is, assigning a keyword to each object or region in...

Descripción completa

Detalles Bibliográficos
Autor: HEIDY MARISOL MARIN CASTRO
Tipo de recurso: tesis de maestría
Estado:Versión aceptada para publicación
Fecha de publicación:2008
País:México
Institución:Instituto Nacional de Astrofísica, Óptica y Electrónica
Repositorio:Repositorio Institucional del INAOE
Idioma:español
OAI Identifier:oai:inaoe.repositorioinstitucional.mx:1009/492
Acceso en línea:http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/492
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/Armar/Assemble
info:eu-repo/classification/Aprendizaje automático/Machine learning
info:eu-repo/classification/Semi-supervisado/Semi-supervised
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/12
info:eu-repo/classification/cti/1203
info:eu-repo/classification/cti/120323
Descripción
Sumario:Content-based image retrieval is a technique that uses the visual content of images such as color, form or texture to search images from large image collections. One way to carry out content-based image retrieval consists of image annotation, that is, assigning a keyword to each object or region in the image and using these keywords to retrieve images. Manual image annotation of a large collection of images is a complex and a subjective task, and it consumes a lot of time. An alternative approach is to automatic annotate images using machine learning techniques. In this thesis, a new algorithm called WSA (Weighted Semi-Supervised Ada-Boost) is proposed for automatic digital image annotation. WSA is based on an assemble of bayesian classifiers using a semi-supervised learning approach. The algorithm WSA uses Naive Bayes as its base classifier. A set of these is combined in a cascade based on the AdaBoost technique. However, when training the ensemble of Bayesian classifiers, it also considers the unlabeled images in each stage. These are annotated based on the classifier from the previous stage, and then used to train the next classifier. The unlabeled instances are weighted according to a confidence measure based on their predicted probability value; while the labeled instances are weighted according to the classifier error, as in standard AdaBoost. The performance of WSA was evaluated using different databases and was compared against other classifiers like NaiveBayes, AdaBoost and the algorithm SA. In the experiments, WSA obtained better performance in the prediction of labels of the images.