Efficient exploration of region hierarchies for semantic segmentation

The problem of object detection has been largely studied in the literature, with many solutions based on sliding windows and interest point analysis. On the other hand, the state of the art on image segmentation has recently achieved important advances, providing more stable solutions that may be ap...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Bellver Bueno, Míriam
Format: Dissertation
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The problem of object detection has been largely studied in the literature, with many solutions based on sliding windows and interest point analysis. On the other hand, the state of the art on image segmentation has recently achieved important advances, providing more stable solutions that may be applied in the object detection problem. This thesis proposal initially focuses on the expansion of the popular deformable parts model for object detection to an image representation based on a hierarchical partition. Such structure can boost the feature extraction with a max-po The motivation of this work is the efficient exploration of hierarchical partitions for semantic segmentation as a method for locating objects in images. While many efforts have been focused on efficient image search in large-scale databases, few works have addressed the problem of locating and recognizing objects efficiently within a given image. My work considers as an input a hierarchical partition of an image that defines a set of regions as candidate locations to contain an object. This approach will be compared to other state of the art algorithms that extract object candidates for an image. The final goal of this work is to semantically segment images efficiently by exploiting the multiscale information provided by a hierarchical partition, maximizing the accuracy of the segmentation when only a very few regions of the partition are analysed. La motivación de este trabajo es explorar eficientemente un árbol jerárquico para segmentar semánticamente imágenes como método para reconocer objetos. Muchos trabajos han tratado la búsqueda eficiente de objetos en imágenes desde el punto de vista global de la imagen en grandes bases de datos, pero no se han dedicado tantos esfuerzos en resolver el problema de localizar y reconocer objetos dentro de la propia imagen. Esta tesis trabaja con particiones jerárquicas de una imagen que definen un conjunto de regiones candidatas para contener un objeto. Segmentar imágenes utilizando estas regiones se comparará con los resultados obtenidos a partir de candidatos de objeto extraídos mediante otros algoritmos del estado del arte. El objetivo final es segmentar imágenes semánticamente de forma eficiente aprovechando la información entre niveles del árbol jerárquico, maximizando la calidad de la segmentación cuando sólo un conjunto muy reducido de zonas del árbol son analizadas. La motivació d'aquest treball és l'exploració eficient d'un arbre jeràrquic