UNSUPERVISED MANIFOLD LEARNING BY CORRELATION GRAPH AND STRONGLY CONNECTED COMPONENTS FOR IMAGE RETRIEVAL

Guimaraes Pedronette, Daniel Carlos [UNESP]; Torres, Ricardo da S.; IEEE

UNSUPERVISED MANIFOLD LEARNING BY CORRELATION GRAPH AND STRONGLY CONNECTED COMPONENTS FOR IMAGE RETRIEVAL

Date

2014-01-01

Authors

Guimaraes Pedronette, Daniel Carlos

Torres, Ricardo da S.

IEEE

Publisher

Ieee

Type

Work presented at event

Access right

Acesso aberto

Abstract

This paper presents a novel manifold learning approach that takes into account the intrinsic dataset geometry. The dataset structure is modeled in terms of a Correlation Graph and analyzed using Strongly Connected Components (SCCs). The proposed manifold learning approach defines a more effective distance among images, used to improve the effectiveness of image retrieval systems. Several experiments were conducted for different image retrieval tasks involving shape, color, and texture descriptors. The proposed approach yields better results in terms of effectiveness than various methods recently proposed in the literature.