$h$-$\gamma$ Blossoming, $h$-$\gamma$ Bernstein Bases, and $h$-$\gamma$ Bézier Curves for Translation Invariant $\left(\gamma_{1},\gamma_{2}\right)$ Spaces

This paper introduces a novel $h$-$\gamma$ blossoming framework for translation invariant $(\gamma_1, \gamma_2)$ spaces by merging $\gamma$-blossoming with $h$-blossoming, and utilizes this framework to define and analyze the properties, algorithms, and geometric characteristics of corresponding $h$-$\gamma$ Bernstein bases and Bézier curves.

The Gist

Imagine you are an architect designing a bridge. Usually, you have two main tools: Polynomials (simple, straight lines and gentle curves) and Trigonometric functions (wavy, sine-wave shapes). For decades, mathematicians have had a special "magic blueprint" called Blossoming that helps them design, smooth, and manipulate these shapes perfectly.

However, there was a problem. This magic blueprint worked great for polynomials and great for waves, but it didn't work well when you tried to mix them or use them in new, discrete (step-by-step) ways. It was like having a wrench that only fits bolts, and a screwdriver that only fits screws, but no tool for the weird, custom-shaped fasteners you needed for a new type of spaceship.

This paper introduces a universal tool called the $h$–$\gamma$ Blossom. It's a new way to design curves that works for any shape generated by two basic building blocks, whether they are straight lines, waves, or even strange, step-by-step digital versions of those waves.

Here is the breakdown of their discovery using simple analogies:

1. The Building Blocks: The "Shape Generators"

Think of a curve as a house built from two types of bricks.

• $\gamma_1$ and $\gamma_2$ are your two brick types.
• In a standard polynomial house, the bricks are 1 (flat) and $x$ (a ramp).
• In a trigonometric house, the bricks are $\cos(x)$ (a wave) and $\sin(x)$ (another wave).
• In a "hyperbolic" house, the bricks are $\cosh(x)$ and $\sinh(x)$ (shapes that look like hanging chains).

The authors realized that if these two brick types have a special property called Translation Invariance, you can build a universal system.

• Translation Invariance is like a magic rule: If you slide your whole house to the left or right (shift it by $h$), the new shape is just a mix of the original bricks. You don't need new, weird bricks; you just rearrange the old ones. This rule holds true for polynomials, waves, and even their "digital" (discrete) cousins.

2. The New Magic: The $h$–$\gamma$ Blossom

The "Blossom" is the secret sauce. In the old days, to find the exact shape of a curve at a specific point, you had to use complex math. The Blossom is a function that takes multiple inputs (like $n$ different points) and tells you the curve's value in a very symmetrical, predictable way.

The authors created a hybrid Blossom (the $h$–$\gamma$ Blossom) that combines:

• $\gamma$-Blossoming: The ability to handle those special brick types (waves, hyperbolic shapes).
• $h$-Blossoming: A "step size" parameter ($h$). Think of $h$ as the zoom level or the grid size.
  • If $h = 0$, you get the smooth, continuous curves we know (like classical Bezier curves).
  • If $h > 0$, you get "discrete" curves, which are like pixelated or stepped versions of the smooth curves. This is crucial for computer graphics and digital signal processing where things aren't perfectly smooth.

The Analogy: Imagine a smooth, flowing river (the curve).

• The $\gamma$-part lets you describe the river whether it's made of water, oil, or honey.
• The $h$-part lets you look at the river through a grid. If the grid is fine ($h$ is small), it looks smooth. If the grid is coarse ($h$ is large), the river looks like a staircase. The new Blossom works perfectly for both views simultaneously.

3. The Toolkit: What Can You Do With It?

Once you have this new Blossom, the authors built a whole construction kit around it:

• $h$–$\gamma$ Bernstein Bases: These are the "standard colors" or "standard weights" you use to paint your curve. Just like mixing red and blue paint to get purple, you mix these basis functions to create any curve you want. The paper gives you the exact recipe for these mixes.
• $h$–$\gamma$ Bézier Curves: These are the actual curves you draw on the screen, controlled by "control points" (like pulling on a rubber band).
• Recursive Algorithms (The "De Casteljau" Method): This is a clever way to calculate the curve. Instead of doing one giant, hard math problem, you break it down into tiny, simple steps (like folding a piece of paper repeatedly). The paper shows how to do this folding for these new, weird shapes.
• Subdivision: This is like zooming in. You can take a curve, split it in half, and get two smaller curves that fit together perfectly. The authors proved that if you keep splitting the curve in half over and over, the jagged lines you get will eventually become a perfect, smooth curve.
• Degree Elevation: This is like taking a sketch and making it more detailed without changing the shape. You can turn a simple curve into a complex one that looks exactly the same but has more control points to tweak later.

4. Why Does This Matter?

Before this paper, if you wanted to design a curve that was a mix of a wave and a step-function (common in digital engineering), you had to reinvent the wheel every time. You couldn't use the standard tools.

This paper provides a universal translator.

• It unifies polynomials, trigonometry, and hyperbolic geometry under one roof.
• It bridges the gap between continuous (smooth) math and discrete (digital/computer) math.
• It gives engineers and computer scientists a single, robust set of rules to design, smooth, and animate shapes, whether they are designing a car body, a sound wave, or a digital animation.

In a nutshell: The authors took the "Lego instructions" for building smooth curves and updated them so you can build with any type of brick, in any size, and whether the structure is smooth or pixelated. They gave us a new, super-powerful ruler for the digital age.

Technical Summary

1. Problem Statement

The paper addresses the need to generalize the powerful tools of Computer-Aided Geometric Design (CAGD)—specifically blossoming, Bernstein bases, and Bézier curves—beyond the classical polynomial setting to more general function spaces.

• Context: Classical Bézier curves rely on polynomial spaces. However, many applications require curves that better fit specific physical or geometric behaviors, such as trigonometric (circles, helices), hyperbolic, or discrete analogues of these functions.
• The Gap: While "γ-blossoming" exists for general $(\gamma_1, \gamma_2)$ spaces and "h-blossoming" exists for polynomial spaces with a step parameter $h$, there was no unified framework that combined these concepts. Specifically, existing methods lacked a shape parameter $h$ that could simultaneously:
  1. Generalize the basis to non-polynomial spaces (trigonometric, hyperbolic, discrete).
  2. Provide interpolation capabilities not present in standard $\gamma$-Bernstein schemes.
  3. Maintain the recursive evaluation properties (like de Casteljau's algorithm) essential for efficient computation.

2. Methodology

The authors develop a unified theoretical framework by merging $\gamma$-blossoming (for $(\gamma_1, \gamma_2)$ spaces) with $h$-blossoming (for $h$-Bernstein bases).

A. Translation Invariant $(\gamma_1, \gamma_2)$ Spaces

The paper defines a space $\pi_n(\gamma_1, \gamma_2)$ of order $n$ spanned by functions of the form $\gamma_1^{n-k}(x)\gamma_2^k(x)$. A key requirement is translation invariance: the functions $\gamma_1(x-h)$ and $\gamma_2(x-h)$ must be expressible as non-singular linear combinations of $\gamma_1(x)$ and $\gamma_2(x)$.

• Examples: Polynomials ($\gamma_1=1, \gamma_2=x$), Trigonometric ($\cos x, \sin x$), Hyperbolic ($\cosh x, \sinh x$), and their discrete analogues.

B. The $h$–$\gamma$ Blossom

The core innovation is the definition of the $h$–$\gamma$ blossom, a symmetric, multilinear function $g((u_1, v_1), \dots, (u_n, v_n); h)$ associated with a curve $G(t)$. It satisfies three axioms:

1. Symmetry: Invariant under permutation of arguments.
2. Multilinearity: Linear in each argument pair.
3. $h$–$\gamma$ Diagonal Property: Unlike standard blossoms where $g(t, \dots, t) = G(t)$, this blossom evaluates to $G(t)$ when the arguments are shifted by $h$:
   $$g(\Gamma(t), \Gamma(t-h), \dots, \Gamma(t-(n-1)h); h) = G(t)$$
   where $\Gamma(t) = (\gamma_1(t), \gamma_2(t))$.

The authors prove the existence and uniqueness of this blossom for translation invariant spaces, provided a specific set of basis functions $\{G_{n,k}(t; h)\}$ are linearly independent (conditions for which are detailed in Appendix A).

C. Recursive Evaluation and Dual Functionals

Using the properties of the blossom, the authors derive:

• Recursive Evaluation Algorithms: Generalizations of the de Casteljau algorithm that compute curve points using a triangular scheme involving the determinant function $d(u, v) = \gamma_1(u)\gamma_2(v) - \gamma_2(u)\gamma_1(v)$.
• Dual Functional Property: The control points (coefficients) of the curve can be extracted directly by evaluating the blossom at specific shifted parameter values.

3. Key Contributions

1. Novel $h$–$\gamma$ Blossom: A new blossom definition that unifies the shape parameter $h$ with general function spaces, enabling interpolation and shape control unavailable in previous $\gamma$-schemes.
2. $h$–$\gamma$ Bernstein Basis Functions: The derivation of basis functions $B_n^k(x, [a, b]; \gamma, h)$ that form a basis for $\pi_n(\gamma_1, \gamma_2)$. These bases satisfy a two-term recurrence relation analogous to classical Bernstein polynomials.
3. $h$–$\gamma$ Bézier Curves: A definition of Bézier curves in these spaces where the control points are uniquely determined by the dual functional property of the blossom.
4. Fundamental Identities:
   • Partition of Unity: Proven for specific cases (polynomials, trigonometric).
   • Marsden Identities: Derived to relate the basis functions to the determinant function $d(t, x)$.
   • Degree Elevation: Formulas for elevating the degree of curves (e.g., from order $n$ to $n+1$ for polynomials, or $n$ to $n+2$ for trigonometric spaces).
   • Subdivision: A recursive midpoint subdivision algorithm that generates control polygons converging to the curve.

4. Key Results

• Interpolation Capability: A significant result is Corollary 5.14, which states that if the interval length is $b = a - nh$ (with $h \neq 0$), the resulting Bézier curve interpolates all control points. This is a unique feature of the $h$-parameter that standard polynomial or $\gamma$-Bernstein schemes do not possess.
• Convergence: Theorem 5.18 and 5.19 prove that the control polygons generated by the recursive midpoint subdivision algorithm converge pointwise and uniformly to the original $h$–$\gamma$ Bézier curve, provided $\gamma_1, \gamma_2 \in C^1$.
• Shift Invariance: The basis functions are shown to be shift-invariant (Theorem 5.3), meaning the shape of the basis functions depends only on the relative interval length, not the absolute position.
• Generalization: The framework successfully recovers known results as special cases:
  • $h=0$ yields standard $\gamma$-Bernstein bases.
  • $\gamma_1=1, \gamma_2=x$ yields $h$-Bernstein polynomials.
  • $\gamma_1=\cos x, \gamma_2=\sin x$ yields trigonometric Bernstein bases.

5. Significance

This work provides a comprehensive and rigorous mathematical foundation for extending CAGD techniques to a broad class of non-polynomial spaces.

• Practical Impact: It allows designers and engineers to construct curves that naturally fit trigonometric or hyperbolic geometries (common in robotics, architecture, and physics) while retaining the intuitive control and computational efficiency (via subdivision and recursive evaluation) of Bézier curves.
• Shape Control: The introduction of the parameter $h$ offers a new degree of freedom for shape manipulation, specifically enabling interpolation of control points within the same framework used for approximation, bridging a gap between interpolation and approximation methods.
• Unification: By treating polynomials, trigonometric, hyperbolic, and discrete functions under a single "Translation Invariant" umbrella, the paper simplifies the theoretical landscape of geometric modeling, suggesting that many distinct algorithms are actually instances of a single, generalized $h$–$\gamma$ theory.