Cone-Traced Supersampling With Subpixel Edge Reconstruction.

Area of Science:

• Computer Graphics
• Image Synthesis
• Computational Geometry

Background:

• Signed distance fields (SDFs) offer infinite level of detail but suffer from aliasing when rendered at finite resolutions.
• Existing antialiasing methods for SDFs are computationally expensive (supersampling) or introduce artifacts (post-processing).

Purpose of the Study:

• To present an efficient and robust spatial antialiasing solution for SDF rendering.
• To complement the sphere tracing algorithm without requiring additional rays or offline pre-filtering.

Main Methods:

• Introduced cone-traced supersampling (CTSS) to perform supersampling along traced rays near surfaces.
• Utilized cone intersections within the view frustum to identify areas with partial visibility (object contours).
• Developed subpixel edge reconstruction (SER) to handle complex pixels with geometric edges in flat regions.

Main Results:

• CTSS effectively reduces aliasing in SDF rendering.
• The method is easily implementable in existing real-time SDF renderers.
• Achieved comparable antialiasing quality to conventional supersampling at a significantly lower computational cost.

Conclusions:

• CTSS provides an efficient and practical alternative for antialiasing SDF-based geometry.
• The combination of CTSS and SER offers robust edge reconstruction and aliasing reduction.
• This approach minimizes shading computations while maintaining high-quality rendering.