Abstarct: Borehole image logs are sonic or resisitivity tools used for rock features or fluid in carbonate reservoirs. Fractures are one of the important features can be detected using image logs. Fractures have very important role in fluid flowing and stability in borehole. Fracture detection is a subjective process. Fracture detection is a complex process because of weak contrast of fractures, variable thickness of fractures, noise and some similar events such as bedding. In this thesis, some techniques are proposed using image processing and pattern recognition for automatic fracture detection and tracing in image logs. A basic and vital step for fracture detection is image segmentation.various segmentation methods applied on image logs. And after validating these methods, the best method has been selected for segmentation. Self organizing map indicated better results rather than other methods. This neural network clusters pixels of image using a competitive algorithm using color feature. After segmentation, some local information and morphological operation used to extract fracture pixels from other pixels. Then some points selected as samples for fracture tracing. For fracture tracing, first samples of each fracture distinguished from each others. Then, fractures traced using radial basis networks. In this thesis, just open natural fractures detected and traced. It has been used FMI image logs for fractures detection and tracing.