Open AccessAn adaptive level set method based on joint estimation dealing with intensity inhomogeneity
Author(s) -
Zhu Jiang,
Zeng Yan,
Li Jianqi,
Tian Shujuan,
Liu Haolin
Publication year - 2021
Publication title -
iet image processing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.401
H-Index - 45
eISSN - 1751-9667
pISSN - 1751-9659
DOI - 10.1049/ipr2.12115
Subject(s) - weighting , computer science , robustness (evolution) , segmentation , intensity (physics) , artificial intelligence , energy (signal processing) , object (grammar) , boundary (topology) , computer vision , image segmentation , energy functional , level set (data structures) , algorithm , pattern recognition (psychology) , mathematics , statistics , optics , medicine , mathematical analysis , biochemistry , chemistry , physics , radiology , gene
Automatic object segmentation has been a challenging task due to intensity inhomogeneity. The traditional way is to eliminate the intensity inhomogeneity, which causes the object to lose useful intensity information. The authors propose an adaptive level set method for the segmentation of intensity inhomogeneous images. Firstly, global and local features are utilised to collaboratively estimate the image, which devotes to compensating for intensity inhomogeneity. The local estimation retains detailed spatial information, and the global estimation mainly contains the regional information of the partitioned object. Then, during the construction of the energy functional, joint estimation is introduced to create the external energy. To acquire the precise location of the boundary, a weighting factor indicated by the gradient is introduced into the internal energy. Finally, after the numerical calculation of the energy functional by additive operator splitting algorithm, this method achieves the desired performance in terms of accuracy and robustness. Experimental results verify this method outperforms the comparative methods and can be applied to many real‐world scenarios.