The overwhelming proliferation of digital images on media sharing webs have triggered the requirement of effective tools to retrieve images of interest using semantic concepts. Due to the semantic gap between low-level visual features and high-level semantic concepts of an image, however, the performances of many existing automatic image annotation algorithms are not so satisfactory. In this paper, a novel image classification scheme, named high order statistics based maximum a posterior.
This method first utilizes high order statistics to measure the triplet-dissimilarity to better describe the relevance among images, then utilizes a maximum of a posterior algorithm with the information of Gaussian Mixture Model and dissimilarity increments distribution to estimate the relevance scores of each tag. Experimental results on a general-purpose image database demonstrate the effectiveness and efficiency of the proposed scheme.