Image Segmentation Using Genetic Algorithm-based Maximum Entropy Single Threshold and Maximum Between-Class Variance Methods

Using MATLAB programming language, we have implemented several genetic algorithm-based image segmentation techniques: maximum entropy single threshold segmentation, maximum entropy double threshold segmentation, two-dimensional maximum entropy single threshold segmentation, and maximum between-class variance (Otsu method) segmentation with genetic algorithm optimization. The implementation includes key MATLAB functions such as ga() for genetic algorithm optimization, entropy calculations for histogram analysis, and custom fitness functions that evaluate threshold criteria. For maximum entropy methods, the code calculates the entropy distribution of image histograms to determine optimal thresholds that maximize information preservation. The 2D maximum entropy approach extends this to two-dimensional histogram spaces for more robust segmentation. For the genetic algorithm Otsu method, the implementation optimizes the between-class variance function, where the algorithm evolves population of threshold candidates through selection, crossover, and mutation operations to find the optimal separation between classes. These algorithms not only perform effective image segmentation but also deliver superior results when processing complex images with uneven lighting or texture variations. Through these implementations, we can better understand and analyze image characteristics, providing valuable references and applications for related research fields. Furthermore, by improving and optimizing the algorithms through parameter tuning, adaptive mutation rates, and elite selection strategies, we can further enhance the accuracy and efficiency of image segmentation, potentially leading to significant breakthroughs and advancements in image processing research and applications.