Journal of Digital Information Management PAN Xin, H. Sagan 14 1 2016 http://dline.info/fpaper/jdim/v14i1/v14i1_2.pdf Digital image clustering algorithms can classify pixels according to their data characteristics without the pre-input of training samples. The number of categories and center point value are difficult to determine because of the large size of pixels and several features of digital image data. This paper proposed a digital image clustering algorithm based on multi-agent center optimization (DICA-MCO). The proposed algorithm established a problem optimization and solving system composed of agents. To achieve fuzzy evaluation, DICA-MCO mapped the digital image clustering problem as a problem of intelligent agent movement in a multidimensional solution space. Results demonstrat that compared with traditional algorithms, DICA-MCO can select the optimal number of categories and value of center points and has high classification accuracy, Kappa coefficient, and classification effect.