Image segmentation is one of the crucial steps in image analysis and image understanding that can be used in numerous applications such as bio-medical, military, robotics, remote sensing, document processing and industrial automation. The goal of image segmentation is to divide an image into different regions, each having certain properties. There are many image segmentation techniques in the literature. However, what is suitable for one image may not be so for another. Neural Networks have provided potential alternatives to the traditional techniques for image processing problems. Also, there has been an increasing use of fuzzy logic theory for image processing applications. In this work, a software system is developed using combination of fuzzy and neural network techniques to implement an image segmentation algorithm based on clustering approach. The image clustering is done with the Kohonen Self-'" Organizing network and the Fuzzy Algorithms for Learning Vector Quantization (FALVQ) methods. Two slight modifications on the learning process of these algorithms has been suggested to overcome the problem of inappropriate weight initialization for the net and having to determine the number of clusters in advance. The classical ISODATA clustering algorithm was implemented for comparison purposes. The system consists of the following phases: In the first phase, set of local statistical features is calculated for each pixel in the image. In the second phase, clustering method is used to group the image pixels in the feature space into clusters. Finally, these clusters are mapped back to the spatial domain to produce a segmentation of an image. The obtained results indicate that the Kohonen and FALVQ algorithms can provide adaptive image segmentation algorithm and produce results subjectively comparable and sometimes better than those obtained from the classical ISODATA algorithm. Also, Experiments have showed that the suggested modifications have improved the performance of the system and give acceptable results.