Earticle

초록

영어

In this paper, a hierarchical nonlocal method for image segmentation is proposed. The method is mainly consisting of two stages. In the first stage, a nonlocal segmentation model based on nonlocal differential operators is used to find a smooth solution. Once the solution is obtained, the segmentation is done by thresholding the solution into different phases in the second stage. The K-means method is used to determine the thresholds. One can obtain any K-phase segmentation ( K ³ 2 ) by choosing ( K - 1 ) thresholds. There is no need to specify the number of segments before finding the solution. Due to the convexity of our proposed model the split-Bregman algorithm is adopted to efficiently solve the minimization problem. When the images are severe intensity inhomogeneity, a nonlocal variation Retinex algorithm is used to preprocessing the given images. The true underlying reflectances of the given images are extracting. Then, these reflectances are used as inputs to the proposed nonlocal model. Experimental results show that our method performs better than many two-phase or multiphase segmentation methods. With the preprocessing of the nonlocal variation Retinex algorithm, our method produce better results and is more efficient than many famous local region-based active contour model (ACM) methods for images with intensity inhomogeneity.

목차

Abstract
 1. Introduction
 2. Two-Stage Segmentation Model
 3. Nonlocal Differential Operators and Nonlocal Variation Retinex
  3.1. Nonlocal Differential Operators
  3.2. Filtered Gradients
  3.3. Nonlocal Variation Retinex
 4. Our Proposed Model
  4.1. Nonlocal Segmentation Model
  4.2. Numerical Implementation
 5. Experimental Results
  5.1. Comparisons with the Two-Stage Segmentation Model
  5.2. Comparisons with the RSF Model, the LGDF Model, and LIF Model
  5.3. Comparisons with the KGC Model for Multiphase Segmentation
 6. Conclusions
 References

키워드

저자

• Mo Yan [ National Laboratory of Radar Signal Processing, Xidian University, Xi’an City, China ]
• Peng-lang Shui [ National Laboratory of Radar Signal Processing, Xidian University, Xi’an City, China ]

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  International Journal of Control and Automation

• 간기

  월간

• pISSN

  2005-4297

• 수록기간

  2008~2016

• 십진분류

  KDC 505 DDC 605