Research On Interactive Visualization

Visualization which transfers data into interactive graphical representation contains two elements essentially, namely representation and interaction. In order to highlight the importance of interaction and make clear separation with general human-computer interaction, researchers proposed terminologies like interactive visualization, interactive visual analysis and related visual analytics.Recently, interactive visualization has been one the most important problems in visualization research. With the development of the simulation data, novel interaction devices and visualization interaction requirements, interactive visualization is also listed as one of the “hotspots” in the visualization community. It is necessary to deal with challenges,like how to provide users with intuitive and efficient interactive visualization techniques and make appropriate use of recently invited interaction device to get better interaction experience of the visualization result.In this dissertation, I focus on the key problems in interactive visualization and try to work out some fundamental challenges in interactive visualization with inspirations from other areas, like human-computer interaction and human perception. The major contributions of this dissertation are as follows.1. This chapter presents the intuitive volume eraser, an interactive volume rendering tool, to aid direct volume data visualization and Transfer Function(TF) design. The system adopts sketch-based editing interface which enables interactive exploring and editing operations in an intuitive and natural manner. The data features can be enlightened interactively as user desired. The editing results are saved faithfully with the What You See Is What You Get(WYSIWYG) scheme. This chapter also provides a coupled transfer function editor for users who are used to the traditional TF editing interface. The result of experiments on various data demonstrates the effectiveness of the proposed intuitive volume eraser. The comparison of user experience also shows that our tool outperforms the state-of-the-art approaches in friendliness and efficacy.2. The current multicomponent volume segmentation and labeling methods are mostly hard to get correct segmentation and labeling results automatically and rely hardly on experts’ aids, which make related volume exploration to be time-consuming, laborious and prone to errors and omissions. To solve this problem, we present a novel volume exploration method driven by admitted model. We first apply Gaussian mixture models(GMM)to segment the raw volume. However, different components with similar value are still mixed. To segment these components further, we make use of region-grown principle to produce a fine-grained part segmentation. To label different parts automatically, we found that it is helpful to take advantage of annotated model. However, it is not straightforward to label segmented volume with geometric model automatically. We propose a volume-model correspondence schema to overcome this intractable challenge. Moreover, it is essential to exploit intuitive interactive methods for interactive exploration, so we also developed practical precise interaction techniques to assist volume exploration.Our experiments with various data and discussion with specialists show that our method provides an efficient and impactful way to explore volume data.3. Interactive visualization has become a valuable tool in visual exploration of scientific data. One prerequisite and fundamental issue is how to infer three-dimensional information through users’ two-dimensional input. Existing approaches commonly build on the hypothesis that user input is precise, which is sometimes invalid because of multiple causes like data noise, limited resolution of display devices and users’ casual input. In this chapter, we propose an alternative effective algorithm for inferring interaction position in scientific data, especially volume data exploration. Our method automatically assists user interaction with the defined saliency. The presented saliency integrates data value, corresponding transfer function and user input. The result saliency implies remarkable regions of raw data as existing methods. Moreover, it reflects the areas of users’ concern. Thirdly,it eliminates the errors from data and device, helping users get the region they focus on.Various experiments have verified that our method can reasonably refine user interaction and effectively help users access interested features.4. This chapter proposes spatiotemporal volume saliency to detect and explore salient regions in time-varying volume data. Based on the center-surround hypothesis that the salient region stands out from its surroundings, we extend the spatial saliency to time domain and introduce temporal volume saliency. It is defined as a center-surround operator on Gaussian-weighted mean attribute gradient between steps in a scale-independent manner. By combing spatial saliency and temporal saliency together, our spatiotemporal volume saliency is effective in detecting changes of salient regions. We demonstrate its utility in this regard by automating transfer function design and selecting key frames for time-varying volume data.5. As the development of supercomputer, the scale of the numerical computation result from supercomputer is becoming larger and larger. Usual visualization systems are not able to handle with such large data effectively; as a result, special researchers could not efficiently analyze their data. With Para View, an open source platform, we developed a large CFD data parallel visualization environment CPVE based on Tian He-1A. CPVE has abundant useful functions for CFD, like data pre-processing, feature extraction, volume rendering, geometric rendering, texture-based visualization, interactive visualization and so on. We built stereo display environment and developed 3D interaction tool. In the end,we apply our environment into datasets with various scale and type.