Imperial College London > Talks@ee.imperial > Featured talks > Linear Expansion of Thresholds: a Tool for Approximating Image

Linear Expansion of Thresholds: a Tool for Approximating Image

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Pier Luigi Dragotti.

A novel approach for building image processing algorithms will be presented: the Linear Expansion of Thresholds (LET). Contrary to the usual processing approaches which consist in approximating an image by optimizing some criterion, we propose to approximate the processing itself using a linear combination of a few basic bricks—- “thresholds”.

The whole adaptivity of LET algorithms is then concealed in the few (linear) coefficients of the representation, which can be optimized using the same criterion as the one chosen in usual approaches (e.g., MAP , sparse regularization, total variation, etc.) or new statistical criteria like Stein’s Unbiased Risk Estimate (SURE). We also investigate iterative LET approximations for refining first-order solutions.

The main advantage of the LET representation is its high implementation efficiency due to a large dimensionality reduction (from the many pixels of an image, to the few coefficients of the LET representation), and due to its linearity, which preserves the convexity and quadraticity of the optimization criterion.

Over the last five years, we have applied this approach with success to several image distortion problems: image denoising/deconvolution (SURE-based criterion), and sparse image restoration (data-term + l1 regularization). In all these applications, the quality of the results either reach, or set a new state-of-the-art, while being substantially faster.

BIO : Thierry Blu graduated (MSc) from École polytechnique (Paris, France) in 1986. Then, in 1988, he graduated (MEng) from École Nationale Supérieure des Télécommunications (Paris, France) from which he also obtained a PhD in electrical engineering (1996). From 1988 to 1998, he worked at France Telecom R&D (now “Orange”) as a research engineer first in Hertzian wave propagation, then in signal processing and videotelephony. In 1998, he joined the Biomedical Imaging Laboratory, recently created by Prof. Michael Unser at EPFL (Lausanne, Switzerland), where he became responsible for the mathematical aspects of image processing in connection with biomedical/biological data. Prof. Blu left EPFL at the end of 2007 to join the Department of Electronic Engineering at the Chinese University of Hong Kong, where he is currently a tenured professor.

Prof. Blu is the recipient of several awards from the IEEE Signal Processing Society. Specifically, two Best Paper Awards (2003 and 2006) for papers on wavelets, and on Finite Rate of Innovation theory and a Young Author Best Paper Award (2009) on SURE -LET interscale image denoising, which was also listed in the Reader’s Choice column of the Signal Processing Magazine (September 2007 and January 2008 issues). Prof. Blu is an IEEE Fellow (2012) and a member of the technical committee Signal Processing, Theory and Methods of the IEEE Signal Processing Society. He is currently one of the editors-in-chief of Sampling Theory in Signal and Image Processing, and an associate editor of EURASIP Journal on Image and Video Processing. He has also been associate editor of the IEEE Transactions on Image Processing (2002—2006), of the IEEE Transactions on Signal Processing (2006—2010), and of Elsevier’s Signal Processing (2008—2011).

This talk is part of the Featured talks series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

Changes to Talks@imperial | Privacy and Publicity