Enhancement of the segmentation process of multi-component images using fusion with Genetic Algorithm

Segmentation is a fundamental stage in image processing since it conditions the quality of interpretation. Fusion in remote sensing has one goal which is to combine different images obtained by different sensors in one image or in different images in order to improve the content and the resolution of these images. In this research, a new fusion method is implemented to improve the segmentation process. This method combines the results of different image segmentation methods or combines the results of the same image segmentation method. The new fusion method consists of a genetic algorithm (ga) to combine each two segmented images based on the stability factor which is measured by the functional model (FM). Experiments conducted on SPOT V, and Landsat 7 ETM+ images proved that this method can provide better results compared to the results obtained from using two different segmentation methods. Two methods ate used in these experiments, self-organizing map and hybrid dynamic genetic algorithm cooperation method (SOM-HGA) and the Iterative Self-Organizing Data (ISODATA) method.