Visualization via direct volume rendering is a very powerful technique to explore and interact with volumetric data. The transfer function (TF) in a volume rendering system assigns optical properties, such as color and transparency, to the data values. An appropriate TF can make a vast difference in quality and content of the rendered image. However, It is difficult to derive such a function automatically or manually as it is much dependent on the semantics of a specific data set,which causes many usability problems during the interaction process. We reviewed important usability issues in the TF specification process, derived an user study in which some of these issues can be studied in a controlled way. The main problems in TF specification that we have identified are the following: 1. Too many degrees of freedom (DOF) in the interaction process. 2. Inappropriate design of the user interface and inadequate information for TF specification. The traditional trial-and-error method is basic scheme for TF specification and provides only minimal visual feedback. Therefore, we consider it as our baseline system. In our experiment, we a priori formulated the following three hypotheses: 1. Data-dependent (more specifically, histogram) information assists the users in TF specification. 2. Data-independent (more specifically, suggested TF) information assists the users in TF specification. 3. A graphical user interface with limited DOFs in the TF control assists the users in TF specification. In the experiment, the visual feedback at any time consists of a single TF, with its available controls and feedback. There were 13 participants in the experiment, five female and eight male persons between 19 and 50 years old. Each participant performed four tasks in each of the five interface conditions. Each task involved a different data set, and required the participants to visualize a predescribed structure within the data as good and fast as possible. The baseline interface with free-style control, referred to as condition 1, consists of parts 1a and 1b in Figure 1. The panel 1b controls the course of the experiment. The part 1a allows the user to manipulate the TF by creating and moving control points of a piecewise linear function along the horizontal and vertical direction within a 2D interaction area. Experimental condition 2 includes data-dependent information, and consists of parts 1 and 2 in Figure 1. A cumulative histogram and free-style TF interface are presented at the same time. Experimental condition 3 includes data-independent information, and consists of parts 1 and 3 in Figure 1. In condition 4, both data-dependent and data-independent information are offered, so that all parts in Figure 1 are presented. The interface for condition 5 is shown in Figure 2. It is a user interface that allows for a number of piecewise-linear TFs without (data-dependent) histogram information. Each kind of TF is represented by a graphical icon. The available TF curves have only few control points and limited DOFs. Figure 1: The user interfaces for experimental conditions 1 (part 1), 2 (part 1+2), 3 (part 1+3) and 4 (part 1+2+3). Figure 2: The user interface for experimental condition 5. The obtained results can be summarized as the following: • There is no evidence that histograms could help to improve user performance. • Additional information about possible/suggested TFs may be useful to the novel users, but could not improve performance either. • Interfaces that restrict the number of DOF of the TF do not improve performance, and are not better appreciated by users.
Unfortunately, ACM prohibits us from displaying non-influential references for this paper.
To see the full reference list, please visit http://dl.acm.org/citation.cfm?id=1080463.