1 Visualization, Summarization and Exploration of Large Collections of Images: State Of The Art Jorge Camargo, Fabio González Bioingenium Research Group National University of Colombia {jecamargom, fagonzalezo}@unal.edu.co Abstract—This paper attempts to provide a comprehensive state-of-the-art of the recent technical achievements in visual- ization of large collections of images. Major recent publications are included in this survey covering different aspects of the research in this area, including visualization, summarization and exploration. In addition, some other related issues such as performance measures and experimental setup are also discussed. Finally, based on existing technology and the demand from real- world applications, a few promising future research directions are suggested. Index Terms—State-of-the-art, information visualization, ex- ploration, summarization, image collection visualization I. I NTRODUCTION Due to the amount of multimedia content generated with different kind of devices and to the easy of publishing in the web, it is necessary to build suitable tools that allow us to manage this information. This generates problems like how to find efficiently and effectively the information needed, and how to extract knowledge from the data. These issues have been extensively studied by Information Retrieval (IR) researchers, but the main focus has been textual data [9]. However, there are still a huge amount of work to do on other kind of non-textual data, such as images. Information visualization techniques [43] are an interesting alternative to address the problem in the case of large collection of images. Information visualization techniques offer ways to reveal hid- den information (complex relationships) in a visual represen- tation and allow users to seek information in a more efficient way [49]. Thanks to the human visual capacity for learning and identifying patterns, visualization is a good alternative to deal with this kind of problems. However, the visualization itself is a hard problem; one of the main challenges is how to find low-dimensional, simple representations that faithfully represent the complete dataset and the relationships among data objects [34]. The majority of existent approaches use a 2D grid layout for visualizing results. Figure 1 shows a screen shot of the result for a query in Google Images. The main problem of this kind of visualization is that it does not make explicit the relationships among the presented images and only a portion of the results is shown to the user. In this paper we review in a detailed way these and other issues, and we review the literature for building an updated state of the art. The rest of this article is organized as follows: Section II, presents a description of the visualization issues addressed in Fig. 1: Typical visualization grid layout using Google Images this paper; in Section III, we present visualization techniques; in Section IV, we present summarization techniques; in Section V, we present exploration techniques; in Section VI, we describe performance measures; in Section VII, some tools are described; in Section VIII, we describe some applications; Finally, we conclude the article in Section IX. II. DEALING WITH LARGE COLLECTIONS OF I MAGES Due to the large amount of visual and multimedia data generated in Internet, health centers, enterprises, research com- munity, and others, it is necessary to build new mechanisms that allow us to access multimedia data sets in an effective and efficient way. We are interested on provide to user new ways to navigate collections of multimedia data, specifically images, such that user can visualize and explore it in an intuitive way. The first natural question is how to visualize an image collection? In the original space images are represented by many dimensions, so how to reduce the dimensionality such that users can visualize an image in a two dimension space? Assume that we have a way for visualizing the image collection: how to display a summary of the entire collection in a computer screen? Once we have a way to visualize and summarize the collection, how we allow users to explore the images in an intuitive way taking into account the similarity among images? Finally, how to evaluate the performance of the techniques used to solve the mentioned issues? These questions are open and are being addressed in some recent