Markov random fields and constrained optimization for textured image segmentation

• Autorzy: Mościńska K.
• Konferencja: Conference on Computer Graphics and Image Processing, GKPO' 98 (5 ; 18-22.05.1998 ; Borki, Poland)
• Języki publikacji: EN

Abstrakty

EN

Classical methods of image segmentation , like discontinuity detection or region growing concepts, are not satisfactory in case in textured images. The alternative is the application of stochastic models like Markov Random Fields (MRF) for image modelling and segmentation. Stochastic model may be described in terms of energy function that should be minimized during a relaxation procedure. Instead of doubly-stochastic model, in which boty the intensity and the label process are modelled by the set of deterministic features. Local texture properties are evaluated using local linear transforms or results from the first order histogram. We measure the disparity between spatial freatures on the basis of the Kolmogorov-Smirnov statistics. Stochastic relaxation algorithms is applied for the minimization of the global energy function. The forbidden label configuration task are given. The examples presented in the paper confirm the usefulness of proposed models and the efficiency of the designed algorithms. Parallel implementation of the constrained optimization can be considered due to the local computation.

Słowa kluczowe

EN

Markov random field; texture; image segmentation; stochastic relaxation; constrained optimization

• Wydawca: Institute of Information Technology of the Warsaw University of Life Sciences – SGGW
• Czasopismo: Machine Graphics and Vision
• Rocznik: 1998
• Tom: Vol. 7, No. 1/2
• Strony: 259--268
• Opis fizyczny: Bibliogr. x poz.

Twórcy

autor

Mościńska K.

• Dept. of Electronics, Silesian Technical University, 44-101 Gliwice, ul. Akademicka 16, Poland