Quadratic Programming for Probabilistic Image Segmentation

We present a general framework for image segmentation based on quadratic programing, i.e. by the minimization of a quadratic regularized energy linearly constrained. In particular, we present a new and general derivation of the Quadratic Makov Measure Field models (QMMFs) that can be understood as a...

Descripción completa

Detalles Bibliográficos
Autor: Mariano Rivera
Tipo de recurso: informe técnico
Estado:Versión publicada
Fecha de publicación:2010
País:México
Institución:Centro de Investigación en Matemáticas
Repositorio:Repositorio Institucional CIMAT
Idioma:inglés
OAI Identifier:oai:cimat.repositorioinstitucional.mx:1008/602
Acceso en línea:http://cimat.repositorioinstitucional.mx/jspui/handle/1008/602
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/MSC/Visón por Computadora
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/12
info:eu-repo/classification/cti/1203
info:eu-repo/classification/cti/120302
Descripción
Sumario:We present a general framework for image segmentation based on quadratic programing, i.e. by the minimization of a quadratic regularized energy linearly constrained. In particular, we present a new and general derivation of the Quadratic Makov Measure Field models (QMMFs) that can be understood as a procedure for regularizing the model preferences (memberships or likelihood) as well as efficient optimization algorithms. In the QMMFs the uncertainty in the computed regularized probability measure field is controlled by penalizing the Gini’s coefficient and hence it affects the convexity of the QP problem. The convex case is reduced to the solution of a positive definite linear system and, for that case, an efficient Gauss–Seidel scheme is presented. On the other hand, we present a efficient projected Gauss-Seidel with a subspace minimization for optimizing the non–convex case. We demonstrate the proposal capabilities by experiments and numerical comparisons with interactive two-class segmentation as well as in the simultaneous estimation of segmentation and (parametric and non-parametric) generative models. This paper has been submitted to IEEE TRANSACTIONS ON IMAGE PROCESSING, December 2009