Reducing bias and inefficiency in the selection algorithm
Proceedings of the Second International Conference on Genetic Algorithms on Genetic algorithms and their application
Pattern recognition: statistical, structural and neural approaches
Pattern recognition: statistical, structural and neural approaches
Adaptation in natural and artificial systems
Adaptation in natural and artificial systems
Digital image processing (2nd ed.)
Digital image processing (2nd ed.)
On the distribution and convergence of feature space in self-organizing maps
Neural Computation
Machine Learning - Special issue on applications of machine learning and the knowledge discovery process
Computer Vision
Self-Organizing Maps
Introduction to Algorithms
A Genetic Algorithm for Image Segmentation
ICIAP '01 Proceedings of the 11th International Conference on Image Analysis and Processing
Practical Genetic Algorithms with CD-ROM
Practical Genetic Algorithms with CD-ROM
ISICA'10 Proceedings of the 5th international conference on Advances in computation and intelligence
| Hi-index | 0.00 |
Image segmentation is an essential process in image analysis. Several methods have been developed to segment multicomponent images and the success of these methods depends on the characteristics of the acquired image and the percentage of imperfections in the process of its acquisition. Many of the segmentation methods are parametric, which means that many parameters need to be computed or provided before the segmentation process, and any method that works on one type of multicomponent image cannot necessarily work on another. In addition, many segmentation methods are supervised, where a priori knowledge is needed, such as the number of classes. To overcome these obstacles, a self-organizing map (SOM), which is an unsupervised nonparametric method, was used to segment four different types of multicomponent images (Landsat, SPOT, IKONOS and CASI), and the results compared to those of a new nonparametric unsupervised genetic algorithm (GA) for image segmentation. To improve the performance of the GA, a hill-climbing process and another random heuristic module were added to escape the local-minima trap and to improve the speed of the GA; the new algorithm is called the hybrid genetic algorithm (HGA). Verification of the results was performed using two different techniques: field verification and the functional model. These verification techniques show that the HGA is more accurate in multicomponent image segmentation than the SOM.