Minimization — Local Methods

  • Stan Z. Li
Part of the Computer Science Workbench book series (WORKBENCH)


After the energy function E(f),including both the functional form and the involved parameters, is given and thus the optimal solution f* = min f E(f) is entirely defined, the remaining problem is to find the solution. It is most desirable to express the solution in closed-form but generally, this is very diffi­cult in vision problems due to the complexity caused by interactions between labels. Therefore, optimal solutions are usually computed by using some it­erative search techniques. This chapter describes techniques for finding local minima and discusses related issues.


Penalty Method Lagrange Multiplier Method Hopfield Neural Network Compatibility Function Label Assignment 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer Japan 2001

Authors and Affiliations

  • Stan Z. Li
    • 1
  1. 1.Beijing Sigma CenterMicrosoft Research ChinaBeijingChina

Personalised recommendations