Two-Level Transfer Functions Using t-SNE for Data Segmentation in Direct Volume Rendering.

Area of Science:

• Computer Graphics
• Scientific Visualization
• Data Analysis

Background:

• Transfer function (TF) design is critical for effective volume rendering and data understanding.
• Multidimensional TFs offer detailed control but are complex to manipulate.
• Low-dimensional TFs are user-friendly but struggle with data segregation.

Purpose of the Study:

• To propose a novel two-level transfer function approach for volume rendering.
• To reduce the dimensionality of complex transfer functions for easier manipulation.
• To maintain high visualization quality while simplifying TF design.

Main Methods:

• Extracting multidimensional TF attributes from volume data.
• Applying t-Stochastic Neighbor Embedding (t-SNE) for dimensionality reduction of TF attributes.
• Integrating a classical 2D TF with the t-SNE TF within the direct volume rendering pipeline.

Main Results:

• The proposed two-level TF effectively reduces the complexity of manipulating multidimensional attributes.
• Visualization quality is maintained at a high level comparable to existing methods.
• Segmentation and rendering of volume data are improved through simplified TF control.

Conclusions:

• The novel two-level transfer function offers an effective solution for simplifying multidimensional TF design in volume rendering.
• This approach balances ease of manipulation with high-quality visualization.
• It enables more intuitive interaction with complex volume datasets.