Improving the convergence analysis of linear subdivision schemes

Nira Dyn,Nir Sharon

2024-05-22

Abstract:This work presents several new results concerning the analysis of the convergence of binary, univariate, and linear subdivision schemes, all related to the {\it contractivity factor} of a convergent scheme. First, we prove that a convergent scheme cannot have a contractivity factor lower than half. Since the lower this factor is, the faster is the convergence of the scheme, schemes with contractivity factor $\frac{1}{2}$, such as those generating spline functions, have optimal convergence rate.

Numerical Analysis

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the convergence analysis of binary, unary, and linear subdivision schemes. Specifically, the author focuses on several new results related to the contractivity factor of convergent subdivision schemes. The following are the main problems and objectives of the paper:

1. **Prove the lower limit of the contractivity factor of convergent subdivision schemes**:
- The author proves that the contractivity factor of a convergent subdivision scheme cannot be lower than 1/2. Since the smaller the contractivity factor is, the faster the convergence speed of the scheme is, the subdivision scheme with a contractivity factor of 1/2 (such as the scheme for generating spline functions) has the optimal convergence rate.

2. **Provide further convergence conditions and insights**:
- The author provides more insights and conditions on the convergence of linear subdivision schemes and shows the application of these insights in improving algorithms to determine the convergence of such subdivision schemes.

3. **Improve existing algorithms**:
- The paper improves existing algorithms to more effectively determine the convergence of a given subdivision scheme. This includes analyzing the properties of subdivision schemes through Laurent polynomial symbols and using these properties to judge whether the scheme converges and its convergence speed.

### Formula Summary

- **Refinement Rule**:
\[
(S_a f)_i = \sum_{j \in \mathbb{Z}} a_{i - 2j} f_j
\]
where \( f \) and \( S_a(f) \) are two real - number sequences defined on the set of integers \( \mathbb{Z} \), and \( a \) is a finite real - number sequence, called the mask of the scheme.

- **Convergence Condition**:
\[
\Delta S_a^L(f) \leq \mu \Delta f
\]
where \( \mu \in (0, 1) \) is the contractivity factor and \( L \in \mathbb{N} \) is the number of contractions.

- **Symbol Representation**:
\[
a(z) = \sum_{i \in \mathbb{Z}} a_i z^i
\]
For a mask with finite support, the corresponding symbol is a Laurent polynomial.

- **Convergence Judgment Conditions**:
\[
\sum_j a_{2j}=\sum_j a_{2j + 1}=1
\]
Therefore, the symbol of the convergent subdivision scheme satisfies:
\[
a(-1) = 0,\quad a(1)=2
\]

- **Key Theorems**:
- **Theorem 3.2**: For all \( L \in \mathbb{N} \), we have:
\[
\| S_q^L \| \geq \frac{1}{2}
\]
- **Corollary 3.3**: The contractivity factor of a convergent unary linear subdivision scheme is not less than 1/2.
- **Theorem 3.6**: If the coefficients of \( q(z) \) are non - negative, then the subdivision scheme \( S_a \) converges, and the contractivity factor \( \mu=\max\{S_e, S_o\}<1 \).
- **Theorem 3.7**: If \( |q(-1)|>1 \), then the subdivision scheme \( S_a \) does not converge.

Through these theories and algorithm improvements, the author hopes to improve the understanding and analysis ability of the convergence of linear subdivision schemes, so as to be better applied in fields such as computer - aided geometric design.