MICS: Multimodal image collection summarization by optimal reconstruction subset selection

This paper presents a new method to automatically select a set of representative images from a larger set of retrieved images for a given query. We define an image collection summary as a subset of images from the collection, which are visually and semantically representative. To build such a summary we propose MICS, a method that fuses two modalities, textual and visual, in a common latent space, and use it to find a subset of images from which the collection visual content could be reconstructed. We conducted experiments on a collection of tagged images and demonstrate the ability of our approach to build summaries with representative visual and semantic content. The initial results show that the proposed method is able to build a meaningful summary that can be integrated in an image collection exploration system.