Technical Program

Paper Detail

Paper:IMDSP-P2.1
Session:Restoration
Time:Tuesday, May 18, 13:00 - 15:00
Presentation: Poster
Topic: Image and Multidimensional Signal Processing: Image and Video Filtering
Title: A NEW INTERPRETATION OF TRANSLATION INVARIANT DENOISING
Authors: Gang Hua; Rice University 
 Michael T. Orchard; Rice University 
Abstract: Translation Invariant (TI) image denoising is a frame denoising method. It outperforms orthogonal wavelet thresholding by averaging a collection of image estimates in different orthogonal bases. In this paper, we propose local models for signal in smooth and edge area. Using these models, TI’s gain can be analyzed. Motivated by the edge geometry and analysis of smooth regions, we design a special way to choose an estimate (accordingly the basis generating it) from the collection at each pixel. In this way, the visual quality improvement of TI can also be explained. Insights drawn from this perspective include: (a) the mechanisms by which TI achieves gain are different in smooth and edge regions, and (b) most gain comes from edge regions. We also point to an improved way of choosing the best basis at each pixel, if the position information of edges is available.
 
           Back


Home -||- Organizing Committee -||- Technical Committee -||- Technical Program -||- Plenaries
Paper Submission -||- Special Sessions -||- ITT -||- Paper Review -||- Exhibits -||- Tutorials
Information -||- Registration -||- Travel Insurance -||- Housing -||- Workshops

©2015 Conference Management Services, Inc. -||- email: webmaster@icassp2004.org -||- Last updated Wednesday, April 07, 2004