Visualization With Artistic Style And Perception Consideration

Information visualization is the science of visualizing complicated data into graphics, mainly based on human vision in understanding the data. It is thus an important question about how to better visualize the data to help people understand it. In recent years, visual perception has received more and more attention in the field of information visualization. Since 2012, the annual IEEE VIS conference has a special perception topics. Visual perception has become a hot research topic after visual layout research, multidimensional data visualization, graph and social network and visual analysis.However, the research on the application of quantitative visual perception theory to visualization is still few. Much of research draws lessons from the application of visual perception in image processing, and introduces them into visualization, such as map visualization. However it is hard to expand to another research hotspot field- social networks which are different with images. Visualization based on the theory of visual perception will be more likely to be perceived because it conforms to the visual features, and style visualization has been proved with a stronger visual impression, so we believe that combining style visualization based on quantitative perception theory is a promising direction.This disseration focuses on the following three key problems:1. Style visualization. We present a visual method on the sketch style for oil painting, a sketch stroke model and sketch style visualization approach. The method can generate realistic visualization with sketch style by modeling the basic elements in the visualization(point, line), especially "overshoot" effect in the sketch. It also enhances the expression of visualized data. Through user experiments on a variety of visual styles, our proposed method demonstrates a strong expressive and visual appeal.2. Visualization optimization based on visual scale. This dissertation presents a graph visualization method for multi-scale optimization based on visual attention theory and its application in visualization. First, we propose three criteria: simplification, hierarchy and distinction. Then, we define a sensing range of the nodes in multi scale, and optimize the visualization of the graph by layout deformation and the perception distance of the line size and color. Finally, the comparison and analysis are made on graph visualization based on perceptual optimization at different scales, and the optimization of color perception is further studied. Through the comparison with the classical graph simplification method, the proposed framework can help people better understand the structure and characteristics of the sensing data and we also found that the use of warm colors is more conducive to perception.3.Style visualization based on visual tracking theory. We present a simulation model of visual tracking process – visual flowing model and apply the sketch style to graph visualization. The model is constructed by computing the saliency of nodes or clusters, and the visual layout is optimized based on the model. The experiments show that the sketch visualization based on visual flow model can enhance the visual perception of users, and also enhance effectiveness and application.We apply the above research to micro-blog diffusion visualization as a platform to analyze the mechanism of diffusion and detect abnormal users.The major contributions of this paper include visualization based on visual perception and in-depth study on several key problems of modeling and the corresponding algorithm, and the concrete applications to practical problems. The experimental analysis shows that the proposed method is helpful to improve the visual display effect, can help people better understand the data. The proposed approach has laid an important foundation and provides a reference for future research.