Indexing color images using color band moments and a relevance feedback
A. M. Mohammed, Abdelhafid Bessaid, Bechar Hassane, Taleb Ahmed Abdelmalik, Hichem Arioui, Rochdi Merzouki, Hadj Ahmed Abbassi · AIP conference proceedings · 2008
In most content based image retrieval systems there is the problem of localizing the spatial information. To resolve this problem we have proposed in this paper to split the picture into horizontal and vertical bands and characterize each band by their statistical moments of each color component. But the satisfaction of low level characteristics don't implicate that the semantic needs are satisfied, so the human interaction with the system is necessary. For this reason we have integrated in our system a relevance feedback loop, on which we have proposed a method which gives us the similarity degree in function of the positive and negative examples which are manually denoted.