skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: NONCONVEX REGULARIZATION FOR SHAPE PRESERVATION

Abstract

The authors show that using a nonconvex penalty term to regularize image reconstruction can substantially improve the preservation of object shapes. The commonly-used total-variation regularization, {integral}|{del}u|, penalizes the length of the object edges. They show that {integral}|{del}u|{sup p}, 0 < p < 1, only penalizes edges of dimension at least 2-p, and thus finite-length edges not at all. We give numerical examples showing the resulting improvement in shape preservation.

Authors:
 [1]
  1. Los Alamos National Laboratory
Publication Date:
Research Org.:
Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
OSTI Identifier:
1000753
Report Number(s):
LA-UR-07-0291
TRN: US201101%%537
DOE Contract Number:
AC52-06NA25396
Resource Type:
Conference
Resource Relation:
Conference: IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING ; 200709 ; SAN ANTONIO
Country of Publication:
United States
Language:
English
Subject:
99; DIMENSIONS; IMAGE PROCESSING; PRESERVATION; SHAPE

Citation Formats

CHARTRAND, RICK. NONCONVEX REGULARIZATION FOR SHAPE PRESERVATION. United States: N. p., 2007. Web.
CHARTRAND, RICK. NONCONVEX REGULARIZATION FOR SHAPE PRESERVATION. United States.
CHARTRAND, RICK. Tue . "NONCONVEX REGULARIZATION FOR SHAPE PRESERVATION". United States. doi:. https://www.osti.gov/servlets/purl/1000753.
@article{osti_1000753,
title = {NONCONVEX REGULARIZATION FOR SHAPE PRESERVATION},
author = {CHARTRAND, RICK},
abstractNote = {The authors show that using a nonconvex penalty term to regularize image reconstruction can substantially improve the preservation of object shapes. The commonly-used total-variation regularization, {integral}|{del}u|, penalizes the length of the object edges. They show that {integral}|{del}u|{sup p}, 0 < p < 1, only penalizes edges of dimension at least 2-p, and thus finite-length edges not at all. We give numerical examples showing the resulting improvement in shape preservation.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = {Tue Jan 16 00:00:00 EST 2007},
month = {Tue Jan 16 00:00:00 EST 2007}
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share:
  • In this talk a duality framework for the problem of optimizing a nonconvex quadratic function over a sphere is discussed. Additional insight is obtained from the apparently new observation that this nonconvex problem is in fact equivalent to a convex problem of the same type, from which known necessary and sufficient conditions readily follow. Based on the duality results, some existing solution procedures are interpreted as in fact solving the dual. The duality relations are also shown to provide a natural framework for sensitivity analysis. Some possibilities for extending these results to the problem of optimizing a nonconvex quadratic functionmore » over multiple spheres, will conclude the presentation.« less
  • We consider the well-known nonlinear programming problem with continuous variables: /it Minimize F(x) = F(x{sub 1}x{sub 2},{hor_ellipsis},x{sub n}) = {Sigma}{sub i}f{sub i}(x{sub i}) /it Subject to {Sigma}{sub i}x{sub i} = X, x{sub i} {ge} 0 where the functions f{sub i} are nonconvex and differentiable, for which no general solution is heretofore available. The known solution methods are restricted to the special case where the functions f{sub i} are convex. We derive the necessary and sufficient conditions for a feasible point x to be a local minimum point of F(x). We use the necessary conditions to formulate an efficient procedure formore » determining a set M of feasible points that includes the set of all local minima of F(x). Global minimization is then achieved by minimizing the objective function over its set of local minimum points. Specific examples illustrate the application of the results in solving problems which cannot be solved by any other analytical criteria. We also show that the assumption of differentiability stipulated on the functions f{sub i} is not essential, and that analogous results can be obtained in terms of the one-sided derivatives at the points where the functions are not differentiable.« less
  • Necessary and sufficient global optimality conditions will be presented for various classes of non-convex constrained optimization problems. A new nonlinear extension of Farkas` lemma plays a crucial role in presenting the optimality conditions in dual forms. Relationships between certain constrained least squares problems and the extended Farkas` lemma for some specially structured nonlinear inequality systems will be discussed. Applications of the optimality conditions to constrained convex maximization problems and to certain multiplicative programming problems will also be given.
  • Since G. Debreu`s work on the general equilibrium theory, a great effort has been made to generalize the economics models as well as to weak the hypotheses made in the original model. Two main problems have been studied. The equilibrium price existence and the relation between a Pareto optimum and an equilibrium price whose study is also an aim of this work. Indeed, from the Economics point of view we establish a result which allow us to associate for each Pareto Optimum of a general nonconvex nontransitive economy a non-zero price such that each consumer and producer satisfied for thismore » price {open_quotes}first-order conditions{close_quotes} involving subgradients of the distance function to the preferences and production sets. The most important difference between the obtained results in this topic is the geometrical concepts associated with the {open_quotes}first-order conditions{close_quotes}. Guesnerie (1975) used the normal cone associated with Dubovickii-Miljutin`s tangent cone concept, whereas Yu (1984), Quinzii (1986), Cornet (1986), Kahn and Vohra (1987) and Bonnisseau and Cornet (1988, 1990) used Clarke`s normal cone. Also, Kahn (1987) established a similar result that Cornet by using the approximate normal cone which is always contained in Clarke`s normal cone. Thus, his formulation gives sharper results. The demonstration of afore mentioned results are based on Hahn-Banach separation theorem or on a nonsmooth version of Karush-Kuhn-Tucker`s theorem. Recently, the authors Jofre and Rivera established in finite dimension this result by using the set-valued map of {open_quotes}approximate subgradients of the distance function{close_quotes} which is smaller than approximate normal cone and it is compact, moreover this set-valued map is always upper semicontinuous.« less