Reducing Ambiguities in Line-Based Density Plots by Image-Space Colorization

Yumeng Xue, Patrick Paetzold, Rebecca Kehlbeck, Bin Chen, Kin Chung Kwan, Yunhai Wang, Oliver Deussen

Introduction

Visualizing large-scale line data—such as stock prices, temperature changes, or GPS trajectories—often results in a cluttered "hairball" that is difficult to interpret. While researchers frequently use density plots (heatmaps for lines) to reduce this clutter, these plots introduce a new set of problems: visual ambiguities.

Traditional density plots aggregate data into a single color scale, which can trigger misleading Gestalt perceptions. This paper identifies three primary "visual confusers" that lead to incorrect data interpretation:

  1. Illusory patterns: High-density regions appear where no single coherent trend exists.
  2. Ambiguous continuation: At intersections, it is impossible to determine which line follows which path.
  3. Disconnected clusters: Related trends appear separate because they are not perfectly adjacent in space.
The three types of visual confusers: (a) Illusory patterns, (b) Ambiguous continuation, and (c) Disconnected clusters.

Methodology: Image-Space Colorization

To resolve these ambiguities, the authors propose a pipeline that shifts clustering from the data-space (clustering the lines themselves) to the image-space (clustering the pixels of the plot). This approach focuses on what the user actually sees rather than the underlying raw data, which is often noisy.

The Pipeline

  1. Feature Extraction: The canvas is divided into bins (pixels). For each bin, the algorithm records a "feature set" consisting of the IDs of every line passing through that pixel.
  2. Hierarchical Clustering: Bins are grouped together if they share a similar set of lines. The authors utilize the Overlap Coefficient instead of standard metrics like Jaccard similarity. This allows high-density "main roads" and lower-density "side streets" to be grouped into the same trend if one set is a subset of the other.
  3. Circular MDS Colorization: To assign colors, Circular Multidimensional Scaling (MDS) maps the high-dimensional clusters onto a 2D hue circle. This ensures that clusters with similar line compositions receive perceptually similar colors.
  4. Line Assignment: Finally, individual lines can be mapped back to these clusters, allowing users to filter raw data based on the discovered visual trends.

Key Results and Case Studies

A user study with 25 participants confirmed that traditional density plots are highly ambiguous; participants often disagreed on the number of trends present but remained "moderately to extremely certain" of their incorrect interpretations.

Comparison of (a) raw line visualization, (b) traditional density plot, and (c) the proposed colored density plot for Beijing taxi trajectories.

Beijing Taxi Trajectories

In a dataset of 10,357 taxis in Beijing, a traditional density plot (Figure b above) suggests a continuous circular route around the city center. However, the proposed colorization method (Figure c) debunks this illusion, revealing that taxis actually cluster in specific local sectors of the city. What appeared to be a single circular trend was actually the overlap of many short, localized trips.

Stock Market Data

When applied to NYSE stock data (2005–2017), the colorization revealed a "disconnected cluster" of stocks newly listed after 2010. This group appeared as a separate high-density trend that was completely invisible in standard monochromatic density plots.

Conclusion

By moving clustering into the image-space and applying perceptually uniform color mapping, this method effectively resolves the Gestalt-driven ambiguities inherent in traditional density plots. While limited by the number of distinguishable hues (roughly seven), the approach provides a significantly clearer view of overlapping trends in complex, noisy datasets.

Summary updated

Abstract

Line-based density plots are used to reduce visual clutter in line charts with a multitude of individual lines. However, these traditional density plots are often perceived ambiguously, which obstructs the user's identification of underlying trends in complex datasets. Thus, we propose a novel image space coloring method for line-based density plots that enhances their interpretability. Our method employs color not only to visually communicate data density but also to highlight similar regions in the plot, allowing users to identify and distinguish trends easily. We achieve this by performing hierarchical clustering based on the lines passing through each region and mapping the identified clusters to the hue circle using circular MDS. Additionally, we propose a heuristic approach to assign each line to the most probable cluster, enabling users to analyze density and individual lines. We motivate our method by conducting a small-scale user study, demonstrating the effectiveness of our method using synthetic and real-world datasets, and providing an interactive online tool for generating colored line-based density plots.
BibTeX
@article{xue2023reducing,
  author = {Yumeng Xue and Patrick Paetzold and Rebecca Kehlbeck and Bin Chen and Kin Chung Kwan and Yunhai Wang and Oliver Deussen},
  title = {{Reducing Ambiguities in Line-Based Density Plots by Image-Space Colorization}},
  journal = {IEEE Trans. Vis. Comput. Graph.},
  year = {2023},
}
Figure