Wavelet Transforms using VTKm
Abstract
These are a set of slides that deal with the topics of wavelet transforms using VTKm. First, wavelets are discussed and detailed, then VTKm is discussed and detailed, then wavelets and VTKm are looked at from a performance comparison, then from an accuracy comparison, and finally lessons learned, conclusion, and what is next. Lessons learned are the following: Launching worklets is expensive; Natural logic of performing 2D wavelet transform: Repeat the same 1D wavelet transform on every row, repeat the same 1D wavelet transform on every column, invoke the 1D wavelet worklet every time: num_rows x num_columns; VTKm approach of performing 2D wavelet transform: Create a worklet for 2D that handles both rows and columns, invoke this new worklet only one time; Fast calculation, but cannot reuse 1D implementations.
 Authors:
 Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
 Publication Date:
 Research Org.:
 Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
 Sponsoring Org.:
 USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR) (SC21)
 OSTI Identifier:
 1329546
 Report Number(s):
 LAUR1627385
 DOE Contract Number:
 AC5206NA25396
 Resource Type:
 Technical Report
 Country of Publication:
 United States
 Language:
 English
 Subject:
 97 MATHEMATICS AND COMPUTING; Computer Science
Citation Formats
Li, Shaomeng, and Sewell, Christopher Meyer. Wavelet Transforms using VTKm. United States: N. p., 2016.
Web. doi:10.2172/1329546.
Li, Shaomeng, & Sewell, Christopher Meyer. Wavelet Transforms using VTKm. United States. doi:10.2172/1329546.
Li, Shaomeng, and Sewell, Christopher Meyer. 2016.
"Wavelet Transforms using VTKm". United States.
doi:10.2172/1329546. https://www.osti.gov/servlets/purl/1329546.
@article{osti_1329546,
title = {Wavelet Transforms using VTKm},
author = {Li, Shaomeng and Sewell, Christopher Meyer},
abstractNote = {These are a set of slides that deal with the topics of wavelet transforms using VTKm. First, wavelets are discussed and detailed, then VTKm is discussed and detailed, then wavelets and VTKm are looked at from a performance comparison, then from an accuracy comparison, and finally lessons learned, conclusion, and what is next. Lessons learned are the following: Launching worklets is expensive; Natural logic of performing 2D wavelet transform: Repeat the same 1D wavelet transform on every row, repeat the same 1D wavelet transform on every column, invoke the 1D wavelet worklet every time: num_rows x num_columns; VTKm approach of performing 2D wavelet transform: Create a worklet for 2D that handles both rows and columns, invoke this new worklet only one time; Fast calculation, but cannot reuse 1D implementations.},
doi = {10.2172/1329546},
journal = {},
number = ,
volume = ,
place = {United States},
year = 2016,
month = 9
}

A wavelet function generated by a specially constructed symmetric scale function has been explored for use in edge detection. Experiments showed that relatively refined edge information was obtained in the coarse resolution levels. An edge detection algorithm based on regularization with spacevarying parameters has been developed, where the values of the parameters are adaptively determined iteratively. A multiscale edge detection algorithm using a first order regularization filter has been developed. It is demonstrated experimentally that the high localization performance of the filter is combined with high detection performance by using a multiscale integration scheme. Time performances have been evaluated formore »

Wavelet transforms as solutions of partial differential equations
This is the final report of a threeyear, Laboratory Directed Research and Development (LDRD) project at Los Alamos National Laboratory (LANL). Wavelet transforms are useful in representing transients whose time and frequency structure reflect the dynamics of an underlying physical system. Speech sound, pressure in turbulent fluid flow, or engine sound in automobiles are excellent candidates for wavelet analysis. This project focused on (1) methods for choosing the parent wavelet for a continuous wavelet transform in pattern recognition applications and (2) the more efficient computation of continuous wavelet transforms by understanding the relationship between discrete wavelet transforms and discretized continuousmore » 
Undecimated Wavelet Transforms for Image Denoising
A few different approaches exist for computing undecimated wavelet transform. In this work we construct three undecimated schemes and evaluate their performance for image noise reduction. We use standard wavelet based denoising techniques and compare the performance of our algorithms with the original undecimated wavelet transform, as well as with the decimated wavelet transform. The experiments we have made show that our algorithms have better noise removal/blurring ratio.