Minimum-Support Interpolating Subdivision Schemes

• Minimum-support interpolating subdivision schemes are efficient processes that exactly reproduce input values using the smallest possible mask for a specific reproduction order.
• They employ algebraic constructions, such as Dubuc–Deslauriers formulas and factorization conditions, to achieve precise polynomial and exponential reproduction in both univariate and multivariate settings.
• Their minimal support leads to enhanced computational efficiency in applications like grid-transfer for multigrid solvers, geometric modeling, and function approximation.

A minimum-support interpolating subdivision scheme is a finitely supported subdivision process that is interpolatory—exactly reproducing input values at integer sites—and achieves this property with the smallest possible mask (or stencil) for a given target of polynomial or exponential reproduction and regularity. These schemes are central to numerical grid-transfer in multigrid, geometric modeling, and function approximation, because minimizing mask support directly impacts computational efficiency and locality.

1. Core Definition and Minimality Principles

An interpolating subdivision scheme is defined by a mask $a$ (in univariate or multivariate settings) satisfying the interpolation property $a(0)=1,\, a(M\alpha)=0$ for $\alpha\neq 0$ (where $M$ is the dilation matrix), and induces a subdivision operator $S_a$. Minimal-support schemes are those for which the support of $a$ is as small as algebraically possible, given constraints of the reproduction order (e.g., precise reproduction of polynomials up to degree $d$), symmetry, regularity, and in vector cases, Hermite or Lagrange conditions.

Minimal support is tightly dictated by the algebraic structure of the reproduction constraints. For example, in the Dubuc–Deslauriers family of univariate interpolatory schemes, reproducing all polynomials of degree at most $d$ requires a support of $d+1$ consecutive integer points. In multivariate and anisotropic contexts, the interplay between dilation, symmetry, and degree yields similar algebraic lower bounds (Charina et al., 2017, Han, 2021).

2. Algebraic Construction and Characterization

Univariate and Multivariate Algebraic Structures

In the univariate, primal case, the minimal-support Dubuc–Deslauriers scheme of order $d$ is supported on $a(0)=1,\, a(M\alpha)=0$0 with $a(0)=1,\, a(M\alpha)=0$1, and mask coefficients are given by explicit, often Pochhammer-based, polynomials. Multivariate minimum-support schemes use tensor or quasi-tensor structures, with symbolic constructions e.g.

$a(0)=1,\, a(M\alpha)=0$2

for $a(0)=1,\, a(M\alpha)=0$3 with $a(0)=1,\, a(M\alpha)=0$4 odd and $a(0)=1,\, a(M\alpha)=0$5, where $a(0)=1,\, a(M\alpha)=0$6 and $a(0)=1,\, a(M\alpha)=0$7 are symbols of univariate Dubuc–Deslauriers schemes (Charina et al., 2017). These constructions reflect the solution of Vandermonde-type linear systems under the reproduction and interpolatory constraints, with uniqueness established via nonsingularity arguments.

For dual-type interpolatory schemes, algebraic characterizations involve frequency-domain (Poisson sum) identities, ensuring compactly supported masks meet the necessary trigonometric polynomial constraints for interpolation and reproduction at sublattices, often leveraging factorization conditions such as $a(0)=1,\, a(M\alpha)=0$8 (Romani et al., 2019).

Support Bounds

Theorems establish that the minimal required support for a scheme interpolating polynomials to degree $a(0)=1,\, a(M\alpha)=0$9 is $\alpha\neq 0$0 in the univariate case, grows to rectangles or norm-balls in the multivariate/anisotropic case depending on the dilation, and, crucially, no strictly smaller support can suffice; otherwise, the necessary system of equations is overdetermined and thus singular (Charina et al., 2017, Han, 2021).

3. Interpolatory Conditions and Polynomial/Exponential Reproduction

For a mask to be both interpolatory and of minimal support, algebraic and spectral conditions must be simultaneously enforced:

• Interpolatory Condition: $\alpha\neq 0$1, $\alpha\neq 0$2 for $\alpha\neq 0$3.
• Sum Rules (Polynomial Reproduction): Constraints on linear-phase moments or, equivalently, reproduction equations for all polynomials (or exponentials in the generalized case) up to a given order.
• Spectral Regularity: Imposed via the Strang–Fix conditions, with mask symbols forced to factor as powers of suitable box spline factors, e.g., $\alpha\neq 0$4, yielding reproduction of all polynomials of degree up to $\alpha\neq 0$5 (Romani et al., 2019, Han, 2021).

For schemes targeting exponential reproduction, recent advances use explicit closed-form constructions based on identities involving Chebyshev and big $\alpha\neq 0$6-Jacobi polynomials, leading to minimal-support odd-symmetric interpolatory masks that exactly reproduce exponential and trigonometric spaces of the form $\alpha\neq 0$7 (Bos et al., 20 Jan 2026).

4. Examples and Constructions Across Scheme Types

Primal and Dual Univariate Schemes

• The binary interpolatory 3-point Dubuc–Deslauriers scheme reproduces all affine functions with minimal support $\alpha\neq 0$8.
• The ternary dual “Cantor” mask for constant reproduction (arity $\alpha\neq 0$9, degree $M$0) is $M$1, supported on four integers (Romani et al., 2019).

Anisotropic and Multivariate Schemes

In the bivariate case with $M$2, the support is exactly

$M$3

with maximal radius in each direction precisely laid out. Explicit 2D masks and their stencils are tabulated (see, e.g., mask examples for $M$4 with $M$5 in (Charina et al., 2017)):

Reproduction Deg.	$M$6	Mask Size	Support Condition
1	1	$M$7	$M$8
3	2	$M$9	$S_a$0
5	3	$S_a$1	$S_a$2

Generalized Hermite and Multivariate Schemes

For Hermite schemes (interpolating both function and derivative values), the support size lower bound in dimension one is $S_a$3 for degree $S_a$4. Constructive existence results guarantee that schemes with exactly this minimal support can be built for any degree, and mask examples are explicitly enumerated for low $S_a$5 (Han, 2021).

Exponential-Reproducing Schemes

Closed-form minimal-support interpolatory schemes for exponential function spaces use rational functions in $S_a$6, and have support exactly $S_a$7 for $S_a$8 (Bos et al., 20 Jan 2026). This saturation of the reproduction-support tradeoff is proven via Chebyshev and $S_a$9-Jacobi identities.

5. Quasi-Stationary and Anisotropic Constructions

Allowing cycles of multiple short masks in quasi-stationary schemes can achieve global $a$0 smoothness with each individual mask supported on significantly smaller stencils (e.g., $a$1-ring or $a$2-ring in 2D meshes), which is not possible in the stationary setting due to sum rule and interpolation requirements. Formally, a quasi-stationary scheme uses a cycle of $a$3 masks $a$4 and the convergence and smoothness analysis is conducted via composed mask symbols and difference operator spectral radii (Lu et al., 2024). Detailed design algorithms balance support minimality, sum rule constraints, interpolation, and smoothness bounds, with worked examples on both $a$5- and $a$6-symmetric meshes.

6. Applications, Regularity, and Computational Significance

Minimum-support interpolatory schemes are fundamental in grid transfer for multigrid solvers, CAGD (e.g., surface modeling), wavelet construction, and numerical PDEs. In multigrid, minimal-support interpolatory grid-transfer operators are shown to be as effective as their (generally larger) pseudo-spline counterparts in terms of convergence rates, but strictly superior in computational efficiency and locality, due to smaller stencils and fewer nonzeros per mask (Charina et al., 2017).

Regularity (Hölder exponent, $a$7 smoothness) for these schemes is quantified via transition (joint spectral radius) analyses or Strang–Fix theory, with mask constructions designed to meet desired regularity thresholds within the minimal-support regime. Schemes achieving higher arity or exploiting dual constructions can simultaneously minimize support and maximize regularity beyond what is possible with standard primal, stepwise-interpolatory masks (Romani et al., 2019).

7. Comparison and Theoretical Guarantees

• Primal vs Dual: Dual interpolatory schemes, unencumbered by stepwise constraints, can reach same or higher polynomial reproduction and regularity for the same or even reduced support compared to classical primal schemes (Romani et al., 2019).
• Support Lower Bounds: Algebraic and Fourier-theoretic proofs demonstrate the necessity of these support bounds; any attempt to reduce mask length further violates the (overdetermined) polynomial or exponential reproduction constraints (Han, 2021, Bos et al., 20 Jan 2026).
• Generalizability: Recent results extend these constructions to multivariate, anisotropic, quasi-stationary, and Hermite settings, with explicit parametrizations for masks and systematic algorithms available for producing minimal-support interpolatory schemes across a broad range of function spaces and geometric configurations.

Minimum-support interpolating subdivision schemes thus represent an optimal point in the reproduction–support–regularity triad, balancing exactness, regularity, and computational tractability across univariate, multivariate, stationary, quasi-stationary, and exponential polynomial settings (Charina et al., 2017, Lu et al., 2024, Romani et al., 2019, Han, 2021, Bos et al., 20 Jan 2026).