Golden interpolation

Ying He, Jincai Chang

Abstract

For the classic aesthetic interpolation problem, we propose an entirely new thought: apply the golden section. We present three examples: the golden step interpolation, the golden piecewise linear interpolation, and the golden curve interpolation, dealing with applications of the golden section in degrees 0, 1, and 2 in the plane. This paper introduces novel concepts like the golden cuspidal hill and the golden domed hill to provide a reference for combining the golden section with interpolation and stimulating more related research.

Introduction: The Quest for Aesthetic Curves

In numerical analysis, interpolation is traditionally viewed as a data-fitting problem. Given a set of points, the goal is to construct a function that passes through them while maintaining "fairness"—usually defined as smoothness by minimizing bending energy. However, in fields such as Computer-Aided Design (CAD), industrial styling, and digital art, mathematical smoothness does not always equate to visual beauty.

This paper addresses the gap between numerical precision and aesthetic harmony by integrating the Golden Section ($\phi \approx 1.618$) directly into interpolation algorithms. By transforming data nodes to satisfy golden proportions, the authors propose a framework for "Golden Interpolation" that prioritizes natural harmony over simple curvature minimization.

Methodology: Node Transformation and Golden Hills

The core innovation lies in a "node transform + traditional interpolation" workflow. Instead of interpolating raw data, the algorithm first adjusts node positions to satisfy golden ratios.

Degrees of Golden Interpolation

  1. Golden Step Interpolation (Degree 0): Horizontal segment lengths in a step function are adjusted so adjacent segments approach the ratio $\phi$.
  2. Golden Piecewise Linear Interpolation (Degree 1): This introduces the "Golden Cuspidal Hill." When connecting points with a break, the peak is placed at the golden section of the horizontal span. This mimics the "host on stage" principle, where a subject is placed at a naturally harmonious focal point rather than the center.
  3. Golden Curve Interpolation (Degree 2): Using quadratic splines, the method creates "Golden Domed Hills." The hilltop (where the tangent is parallel to the chord) is forced to align with the golden section of the interval.
Comparison of standard (left) vs. golden (right) interpolation for step functions and 3D stump layouts.

To quantify these aesthetic properties, the authors define metrics such as "Left/Right Golden Error" and "Mixed Golden Error." These allow the subjective quality of beauty to be treated as an objective optimization target that a computer can solve.

Key Insights: Proportion Over Smoothness

Standard splines prioritize minimizing the rate of change in curvature. Golden interpolation suggests that human perception of beauty is more closely tied to the relative proportions of a curve’s features. The "Host on Stage" analogy illustrates this: a performer in the exact center looks static, while one at the golden section looks balanced and dynamic. This principle is applied to the peaks of interpolated curves to create more visually engaging designs.

Results and Applications

The effectiveness of Golden Interpolation is demonstrated through synthetic data and practical design scenarios:

  • Landscape Architecture: As shown in the figure above, the layout of cylindrical stumps in a park feels more harmonious when diameters and heights follow golden proportions compared to standard linear distributions.
  • Industrial Design: In designing 3D vases and headboards, "Golden" versions were qualitatively judged as more balanced and pleasant than those generated by traditional quadratic splines.
  • Artistic Rendering: The method was applied to "meteorite tracks" in digital art, creating a more dynamic distribution of light streaks in the night sky.

Limitations and Future Directions

While visually promising, several challenges remain. The definition of "beauty" is inherently subjective, necessitating future formal user studies to validate whether people actually prefer golden curves over traditional ones. Computationally, solving for "Golden Error" is more expensive than standard interpolation and can introduce larger errors if node additions are not carefully managed. Finally, extending these 1D principles to parametric curves and 3D surfaces is essential for widespread adoption in complex modeling software.

Summary updated

Abstract

For the classic aesthetic interpolation problem, we propose an entirely new thought: apply the golden section. For how to apply the golden section to interpolation methods, we present three examples: the golden step interpolation, the golden piecewise linear interpolation and the golden curve interpolation, which respectively deal with the applications of golden section in the interpolation of degree 0, 1, and 2 in the plane. In each example, we present our basic ideas, the specific methods, comparative examples and applications, and relevant criteria. And it is worth mentioning that for aesthetics, we propose two novel concepts: the golden cuspidal hill and the golden domed hill. This paper aims to provide the reference for the combination of golden section and interpolation, and stimulate more and better related researches.
BibTeX
@misc{he2018golden,
  author = {Ying He and Jincai Chang},
  title = {{Golden interpolation}},
  year = {2018},
  eprint = {1806.09058},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1806.09058},
}
Figure