Interactive important social character identification from large photo collections

In the paper, we describe a mechanism to identify important social characters by analyzing the social structures embedded in photo collections. We first construct a weighted undirected graph from photo collections by examining the co-appearance of individuals in photos, wherein the weights of edges are measures of the social closeness of the involved individuals (vertices in the graph). Then a graph clustering algorithm that maximizes the modularity of the graph partition is applied to detect the embedded social clusters. Once the social clusters are identified, we measure individual's contribution to the formation of a social cluster to quantify the social importance of each character in the cluster. To compensate the discrepancy between the user-perceived important social characters and the algorithmically computed ones, we propose an interactive browsing scheme to enable viewers quickly identify import social characters that conform to their subjectivity. The effectiveness of the proposed mechanism is demonstrated through experiments on consumer photo collections.