Some of the most challenging pattern recognition problems occur when the data is in the form of pictures. Image processing involves the manipulation of image data for viewing by people. Segmentation of an image into homogeneous regions is one of the many intriguing topics In image processing. Clustering technique is a common technique of image segmentation. Many methods for clusters detection have been proposed for an extremely large number of applications. Two of the most important clustering techniques are described in this thesis. In the recent years, enough research initiatives are directed toward exploring evolutionary class of algorithms that are primarily used for search optimization. Genetic Algorithms (GAs) are instances of such evolutionary classes. In our work, a genetic algorithm has been investigated and used to find optimal clusters representation to segment an image. We modified a new crossover operator and implemented it to the algorithm, then we applied this algorithm on some images, which gave superior results. Afterward, the resulting solution had been refined by constructing K-mean algorithm.