Institutt for informatikk
Interview | Visualization Group

3D Visualization World interviewed Dr. Helwig Hauser: Advanced Visualization Research Explores New Paths

Advanced visualization is exploring new paths to reveal the function and structure of relevant natural phenomena or other artifacts of interest. Medicine, geoscience, and engineering are just three selected areas where new visualization research is needed. 3D Visualization World interviewed Dr. Helwig Hauser, professor in visualization at the Department of Informatics at the University of Bergen, Norway to learn more about current research into these areas, and to learn more about some of the visualization activity within his research group.

Green PARP representations in the attached figure were partially flattened in order to reveal the branching structure as-good-as-possible, while the gray representations show simple projections from 3D to 2D.
Traditionally, visualization was concerned with the (interactive) depiction of individual phenomena/artifacts---nowadays, we need to visualize extensive sets, for example, when visualizing entire cohort studies or ensemble datasets. New solutions are needed---for instance to provide the user with an effective comparative visualization. When visualizing spatial phenomena/artifacts (for example PARP models as in the attached figure), a carefully designed partial spatial abstraction can help the user to see, what she or he is interested in (the green PARP representations in the attached figure were partially flattened in order to reveal the branching structure as-good-as-possible, while the gray representations show simple projections from 3D to 2D). Just recently, we have worked out a formal model for comparative visualization of spatial data ensembles, which enables us to support the visualization design with semi-automatic optimization---this way, the user can quickly browse through multiple versions of comparative visualization---each of which optimized according to a particular interest of the user.
I. Kolesar; S. Bruckner; I. Viola; H. Hauser, "A Fractional Cartesian Composition Model for Semi-spatial Comparative Visualization Design," in IEEE Transactions on Visualization and Computer Graphics


3DVW: Could you please explain how you became interested in visualization? What sparked your interest and the direction that you took in developing your career?

HH: Something graphical is in my family, I guess, with artists as my mother and my grandfather (and I’m very interested in the fine arts, myself). At the end of my computer science studies, I got involved with computer graphics, which I liked a lot. From there, it wasn’t a large step toward visualization. Instead of mimicking the real world (with photo-realistic computer graphics), I was intrigued by the idea to show “more than what the real world can do”, for example, to enable a doctor to look into a patient. I have had the pleasure to collaborate with colleagues from different domains and I am more convinced than ever before that visualization is key to successfully connecting users to their data.

3DVW: In reading about your research group you state: “The UiB (University of Bergen) Visualization Group focuses on researching and teaching of new solutions for the efficient and effective visualization of large and complex datasets...for the purpose of data exploration, analysis, and presentation.” Can you explain why large and complex datasets require these new solutions? What is so unique about them to need such special attention?

HH: If data is small and simple, no visualization is needed, usually—a simple look at the data in their raw format will do. If data becomes large, visualization can help, particularly where the size of the data is in a range between small and huge. But many users have also big problems with comparably small data (as Christian Chabot, CEO of Tableau, said in 2008), in particular, when the data is complex (heterogeneous, multi-dimensional, highly structured, multi-modal, etc.). Sometimes, when people talk about Big Data, they simply think about the extended size of data, which is challenging (certainly a big challenge!)—the variety and velocity aspects of Big Data (compare to Gartner’s definition of Big Data), however, are at least equally important and require advanced visualization solutions to enable users drawing insight from such data.

3DVW: Why is 3D important to your work? Could you explain using a few examples please.

HH: In the research of new visualization solutions, 3D clearly plays an important role. Mostly, this is the case, when the data is “somehow” 3D, for example, when the data represents spatial phenomena/artifacts (patients in medicine, fluid flows in engineering, geology, etc.). Occasionally, we also choose 3D as the space in which we construct a visualization of data without an immediate connection to the three spatial dimensions of our world. An example would be a graph embedding problem, in which we aim at preserving inter-node distances as accurately as possible (and this, depending on which graph is given, can be achieved much better in higher dimensions than in lower, for example).

3DVW: We often have readers who have been collecting high quality 3D geodata sets for marine, natural resources and engineering purposes for many years, but they have never really invested funding or time into evaluating this data – using only pieces of it. How valuable is old or historical data in terms of visualization and gaining new knowledge for current purposes?

HH: There may be many reasons for not finding any good use of old data, actually—it’s assumed that much of the data that previously was acquired will remain unused. Still, in certain cases, it can be very valuable to having “old” data in addition to current data, for example, when aiming at prediction. In climatology, for example, any “old” data that contributes to better describing the climate in the past also helps to improve the predictions of climate change in the future. In such cases, however, we usually meet a data integration challenge, since often data formats, data resolution, etc., are varying substantially over time—in particular, when we aim at using “very old” data.

3DVW: A significant amount of visualization activity arises through medical innovations and research. Do these applications and research have broader applicability for marine, climatology, seismic and others kinds of applications?

HH: Medical visualization is maybe the most prominent application domain for our basic research in visualization—most of our publications, at least, relate to medicine in order to demonstrate the applicability of the new solutions, which we research. It is usual, however, that many of the ideas, which at first aim at medicine in terms of the target application, then also prove their usefulness (often in a slightly adapted form) in other domains, including engineering, geosciences, etc. 3D visualization techniques, for example, that are based upon volume rendering, or the like, are a prime example, but also other solutions, including space deformations for visualization, have a broader applicability.

3DVW: Readers often voice concerns about having results for their data, but lacking the understanding and means to represent and illustrate the results effectively. Could you describe some of the considerations for illustrating and presenting results from big data sets?

HH: Illustrative visualization is one of the research branches in our work. In this area, we are aiming at an improved representation of complex phenomena/artifacts—often for documentation, dissemination, and/or teaching purposes. We learn from professional illustrators, for example, medical illustrators, who learned effective forms of illustration, not at the least 3D illustration. Recently, we have integrated some of these concepts, for example, colored shadows, contour lines, etc., into interactive 3D visualization solutions. With such a solution, not only static illustrations are possible, but interactive 3D illustrations, which still carry the expressiveness of professional illustration.

3DVW: What do you consider to be the major challenges facing 3D and visualization of complex and large data sets today?

HH: In my opinion, still the following, principle challenge is most important: Our own human visual system is not really made for 3D vision—our eyes are relatively close to each other and they are looking basically into the same direction (not good for stereo vision). Even though our visual system has been equipped with a number of mechanisms to reconstruct 3D vision in our brain (interpreting motion parallax and occlusion, for ex.), we simply aren’t in a good situation with 3D perception—quite some effort is needed to make 3D work (interaction, advanced techniques supporting depth perception, etc.).

In visualization, to also address this point, I think that the following challenge is really crucial: Visualization is in principal expensive, because it requires the user’s time (as compared to automated solutions), so it really is important to think about the cost-benefit relationship for any visualization solution (and, of course, also for others). When it comes to data exploration and analysis, we recently started to look into a research direction, which we call Visual Analytics. There, one idea is to aim at combined solutions, where parts of the task are solved in an automated fashion and only those parts of the task are solved interactively, where the benefit outweighs the costs.

3DVW: Please tell us about some of the research that you are involved in at the present time and why it is important.

HH: We experience a transition point from traditional visualization to new visualization challenges with respect to several different aspects: We see the need to visualize function in addition to structure (for example: physiology in addition to anatomy), populations in addition to individuals (for example: an entire patient cohort in addition to “just” one patient), and multi-scale perspectives in addition to single-scale considerations (for example: integrating molecular biology in precision medicine), to mention three selected new research challenges of current interest. Also, we wish to integrate heterogeneous data subsets (for instance, from different sensors) and enable the joint exploration and analysis with interactive visualization. Swift hypothesis generation, for example, is a promising opportunity, in particular, when considering rich data with hundreds of dimensions—often also of very different nature (spatial vs. non-spatial, for example).

3DVW: Is automation playing any role in the visualization of complex data today? Can you describe a few examples?

HH: As mentioned, an important part of our research is angled towards combined methods for data exploration and analysis (combined in the sense of combining automated, computational data analysis methods, for ex., from machine learning, statistics, etc., with interactive, visual methods, i.e., visualization). Examples include: an informed strategy to dimension reduction, semi-automatic clustering of long time-sequences or optimization of accurately defined sub-tasks, etc.

3DVW: We understand that the 2nd Virtual Geoscience Conference (VGC 2016) will be held in Bergen, Norway shortly. Can you tell is what that is about and some of the topics included at the event – who should attend?

HH: The VGC 2016 conference (in Bergen, Norway, about one week ago) brought together over 100 experts in different fields of geosciences, who find it promising (if not even necessary!) to consider computational tools (for example for data acquisition, exploration, and analysis) for their research. The conference provided impressive views on new ways to acquire rich geoscientific data (high resolution, multi-variate, time-dependent, etc.) and how to make sense out this data. The panel at the end of the conference concluded that this development is just in its beginning and that an impressive growth of technology-supported geoscience is to be expected within only short time. It is highly likely that an even larger VGC conference #3 will be organized relatively soon.