Image Retrieval Based on Dynamic Local Color Histogram
YING Long,WU Ling-da,LUAN Xi-dao
DOI: https://doi.org/10.3321/j.issn:1000-6788.2006.11.013
2006-01-01
Systems Engineering - Theory & Practice
Abstract:The color histogram based image retrieval method doesn't use the spatial distribution information of the image;Image regular segmentation or extracting some main objects,then compute local color histogram will arouse error.In this paper, image is irregularly segmented by Mean shift clustering.Based on image segmentation,one dynamic local histogram is defined,one statistic which can denote image color and spatial distribution information is designed,one image similarity computing method is also represented.This algorithm is better then the local histogram method which using the image location information,the rotation inflexibility and scale inflexibility of the global color histogram method is also reserved.The experiment result show that using image spatial distribution information improve the accuracy of image retrieval,those images color content are similar but the spatial distribution are different will not appear,this algorithm is efficient and robust.