Image Indexing Using Shape Feature
Qian Zhao, Hui-li Guo · 2011
Based on the analysis of the traditional image retrieval in DCT compressed domain, a novel extraction algorithm is proposed in this paper. The DCT block is firstly classified into different edge type by exploiting five AC coefficients. Then, the edge spatial distribution map is constructed. According to the statistical feature of the map, the edge spatial distribution feature (ESDF) is presented. Meantime, the effect of the energy of DCT blocks on retrieval were considered, a weight is introduced into ESDF. The experiment results show that the new method has a good performance both in retrieval efficiency and effectiveness.