Intensity-Based Image Segmentation
Omer Demirkaya, Musa H. Asyali, Prasanna K. Sahoo · 2008
CONTENTS 6.1 Introduction ........................................................................................... 223 6.2 Between-Class Variance ....................................................................... 224 6.2.1 Criterion Functions Equivalent to BCV ................................. 228 6.2.2 BCV as an Image Bimodality Measure .................................. 231 6.2.2.1 Bimodality Threshold for Uniform Distribution ................................................................. 232 6.2.2.2 Bimodality Threshold for Normal Distribution .... 232 6.2.3 An Iterative Implementation of BCV for Trimodal Images ......................................................................................... 233 6.3 Minimum Error Thresholding ............................................................ 236 6.4 Extention to Multithresholding .......................................................... 241 6.5 Entropy-Based Thresholding .............................................................. 241 6.5.1 One-Dimensional Entropy-Based Method ............................ 242 6.5.2 Two-Dimensional Entropy-Based Method ............................ 246 6.6 Image Segmentation by K-Means Clustering ................................... 258 6.7 Image Segmentation by Fuzzy C-Means Clustering ....................... 260 6.8 Mixture-Modeling-Based Segmentation ........................................... 262 Problems ...........................................................................................................274 References ........................................................................................................ 276 Image segmentation using histogram-based thresholding is probably the most common approach, since it is easy to implement and requires less CPU resources to run. These methods generally employ the maximization or minimization of a criterion function based on the image histogram. The optimal threshold is the gray-level intensity at which the criterion function attains its extremum (maximum or minimum). Thresholding methods are called global if a single threshold is calculated for the entire image. If the image is divided into sub-blocks and a threshold is calculated for each sub-block, then this method of thresholding is called local thresholding.