Edge region color autocorrelogram: A new low-level feature applied in CBIR

Simple yet effective feature extraction is crucial for content- based image retrieval (CBIR). In this paper we propose a novel type of regional feature, which is called edge region color autocorrelogram (ERCAC). It aims to combine the color and shape characteristics of image jointly, by capturing both color distribution of image and spatial correlation of edge points with a structure-based early fusion. Hence, both color information and sketch information are encoded into a single representation. Experimental results show that our method has better performance on the task of TRECVID 2005 concept detection.