Reconstructing Refinable Functions

Nathanael Berglnd
In the summer of 2001 for a job I did an REU (Research Experience for Undergraduates) at Texas A&M University. The topic for the summer was Matrices and Wavelets. One of the fundamental building blocks of wavelets is a function called the scaling function. This function satisfies an equation known as a refinement equation. The set of all functions which satisfy a refinement equation is known as the set of refinable functions. These were the functions I focused my study on. Some of the questions asked included: Can you find a way to determine the class of all refinable functions? all continuous refinable functions? all refinable functions with compact support? etc. At the end of the summer I had found a classification theorem that answered these questions. I also wrote a software program to construct refinable functions based on the theorem.
A refinable function is a function which is equal to the sum of scaled copies of itself, where the copies are contracted horizontally by 1/2, shifted left or right by multiples of 1/2 and scaled by real numbers. For example, the function on the left can be refined as: f(x) = f(2x) + f(2x - 1)
The function below is known as the "hat" function. Notice that it can be refined as: f(x) = 1/2 f(2x+1) + f(2x) + 1/2 f(2x-1)
This function can be obtained from varying the coefficients between the two above functions, so that f(x) = (1-λ) f(2x + 1) + f(2x) + λ f(2x - 1), where 0 ≤ λ ≤ 1. It is possible to show that this function is continuous.
The function on the left shows what happens if you allow λ to grow beyond 1. Note that this is no longer continuous. It is, however, still refinable over a set known as the dyadic numbers. Whether or not this function can be extended to all real numbers is a question I have not yet answered.

Below is a function known as the Daubechies scaling function, so named because it was discovered by a mathematician named Ingrid Daubechies. It's refinement coefficients are:
Here are some more refinable functions I was able to discover. Can you guess what their refinement is?
 

For more information about the results I discovered, or to obtain software, see the links below. To better understand the software and what it does, I highly recommend you read the summary of results before running any of the programs.
summary.pdf - A summary of my results.
software.html - Describes the programs I have written.
paper.dvi - The full length paper detailing my results.

Home Class Schedule Java Maze Mobius Stip One-Sided Surfaces Refinable Functions Rubik's Hypercube Spacetime


DISCLAIMER: THIS PAGE IS NOT A PUBLICATION OF THE GEORGIA INSTITUTE OF TECHNOLOGY AND THE GEORGIA INSTITUTE OF TECHNOLOGY HAS NOT EDITED OR EXAMINED THE CONTENT. THE AUTHOR(S) OF THE PAGE ARE SOLELY RESPONSIBLE FOR THE CONTENT.