Unsupervised manifold learning by correlation graph and strongly connected components for image retrieval

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