Image Collection Summarization With Intrinsic Properties
xmlui.mirage2.itemSummaryView.MetaDataShow full item record
Visual summarization is an inherently complex process as its definition has some subjectivity in itself that there is nothing like a single perfect summary. In general, a good summary consists of two main properties which are (i) coverage and (ii) diversity. A good summary should have a high coverage, i.e it should consist of the key events and the concepts for a given set. At the same time, a good summary should also be diverse, i.e it should not consist of similar events and concepts. In addition to these two main properties, intrinsic image properties such as their aesthetics, their popularity, their sentiment, etc. are assumed to be important especially for the social media applications. In this thesis, we propose an automatic summarization method which considers intrinsic properties of images in addition to coverage and diversity for personal image collection summarization. To evaluate the proposed method, we collected two benchmark datasets where the ground truth summaries are obtained via crowdsourcing. Our experimental analysis reveals that taking intrinsic properties into account improves the summarization performances.