GLI-Color
Salah Bougueroua, Bachir Boucheham · 2015
In this paper, we propose a novel method for color features based image retrieval. The proposal exploits the advantage of local histograms using a distinguished design to delimiting the concerned regions through an elliptic form, organized in circular-concentric manner of increasing size. For each elliptic region, normalized color histograms in the RGB color space are established. Through this sophisticated design, our method will gradually integrate the locality information, thus it is called: Gradual Locality Integration of Color features (GLI-Color). The experiments are conducted on Corel-1k database, using Manhattan and d1 distance metrics. The comparisons of obtained results with global histogram and some published works have shown a significant higher performance.