Multiscale texture segmentation using wavelet-domain hidden Markov models

Author

Choi, Hyeokho ; Baraniuk, Richard

Author_Institution

Dept. of Electr. & Comput. Eng., Rice Univ., Houston, TX, USA

Volume

2

fYear

1998

fDate

1-4 Nov. 1998

Firstpage

1692

Abstract

Wavelet-domain hidden Markov tree (HMT) models are powerful tools for modeling the statistical properties of wavelet transforms. By characterizing the joint statistics of the wavelet coefficients, HMTs efficiently capture the characteristics of a large class of real-world signals and images. In this paper, we apply this multiscale statistical description to the texture segmentation problem. Using the inherent tree structure of the HMT, we classify textures at various scales and then fuse these decisions into a reliable pixel-by-pixel segmentation.

Keywords

hidden Markov models; image classification; image segmentation; image texture; statistical analysis; trees (mathematics); wavelet transforms; image segmentation; joint statistics; modeling; multiscale texture segmentation; pixel-by-pixel segmentation; real-world signals; statistical properties; texture classification; wavelet coefficients; wavelet transforms; wavelet-domain hidden Markov tree models; Classification tree analysis; Discrete wavelet transforms; Hidden Markov models; Image segmentation; Pixel; Statistics; Time measurement; Tree data structures; Wavelet coefficients; Wavelet transforms;

fLanguage

English

Publisher

ieee

Conference_Titel

Signals, Systems & Computers, 1998. Conference Record of the Thirty-Second Asilomar Conference on

Conference_Location

Pacific Grove, CA, USA

ISSN

1058-6393

Print_ISBN

0-7803-5148-7

Type

conf

DOI

10.1109/ACSSC.1998.751614

Filename

751614