This thesis investigates the possibility of generating templates for pattern matchingoptical character recognition systems. Utilizing a quantitative and experimental approach we developed a program that takes input images of a water meter and generates numerical character templates. It uses a series of pre-processing tools, region of interest identification, image cropping and template identification to achieve this. The resulting templates are then compared against manually created templates of the same input images. This is done with the use of an optical character recognition algorithm that has been developed by the authors of this thesis. The results of each step of the process are shown with images comparing the image before and after a processing step has been taken. Along with the graphs measuring the results of the optical character recognition algorithm.While the generated templates generally have a poorer result than the manually created ones. It is deemed that that process was a success. This is due to the fact that there are areas were improvements can be identified quite easily. The authors of this thesis believe that if these improvements were to be made the results of the generated templates would be increased to a satisfactory level.