Ming-Jun Lai,David W. Roach
Abstract:In this paper, we give a parameterization of the class of bivariate symmetric orthonormal scaling functions with filter size $6\times 6$ using the standard dilation matrix 2I. In addition, we give two families of refinable functions which are not orthonormal but have associated tight frames. Finally, we show that the class of bivariate symmetric scaling functions with filter size $8\times 8$ can not have two or more vanishing moments.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to construct bivariate symmetric orthogonal wavelet bases with specific properties, especially those wavelet bases with short support (i.e., finite length). Specifically, the authors focus on how to parameterize bivariate symmetric orthogonal scaling functions that satisfy the following conditions:
1. **Orthogonality**: The scaling function and its translated versions are orthogonal at different positions.
2. **Symmetry**: The scaling function and its corresponding filter coefficients are symmetric, which ensures the linear - phase property.
3. **Vanishing Moments**: The scaling function has certain vanishing moments, so that it can better represent polynomial signals.
In addition, the paper also explores two additional problems:
- Construct some refinable functions that are non - orthogonal but have tight - frame properties.
- Prove that for 8×8 filters, there are no bivariate symmetric scaling functions with two or more vanishing moments.
### Main Problem Decomposition
1. **Parameterization of 6×6 Filters**:
- The paper first constructs a complete parameterization form of all trigonometric polynomials \( m(x,y) \) that satisfy the above conditions for 6×6 - sized filters (i.e., \( N = 5 \)).
- This includes finding a two - parameter family that contains trigonometric polynomials related to the scaling function and showing the specific expressions of these parameterization forms.
2. **Functions that are Non - orthogonal but Have Tight - frame Properties**:
- For other trigonometric polynomials that do not belong to the above two - parameter family, the paper shows that although they are not orthogonal, some tight - frame families can be determined.
3. **Proof of Impossibility for 8×8 Filters**:
- Finally, the paper proves that for 8×8 - sized filters (i.e., \( N = 7 \)), there are no bivariate symmetric scaling functions with two or more vanishing moments. This conclusion is obtained through strict mathematical derivations.
### Markdown - format Display of Formulas
To ensure the correctness and readability of the formulas, the following are several key formulas displayed in Markdown format:
1. **Orthogonality Condition**:
\[
|m(x,y)|^2+|m(-x,y)|^2+|m(x,-y)|^2+|m(-x,-y)|^2 = 1
\]
2. **Symmetry Condition**:
\[
m\left(\frac{1}{x},\frac{1}{y}\right)=x^{-N}y^{-N}m(x,y)
\]
3. **Vanishing Moments Condition**:
\[
m(x,y)=(x + 1)^M(y + 1)^M\tilde{m}(x,y)
\]
where \(\tilde{m}(x,y)\) is another trigonometric polynomial.
4. **Parameterization Form of 6×6 Filters**:
\[
m(x,y)=f_a(x^2,y^2)+xf_b(x^2,y^2)+yf_c(x^2,y^2)+xyf_d(x^2,y^2)
\]
Through the solution of these problems, the paper provides new tools and theoretical foundations for image processing and other application fields.