Optical fiber sustains scratches, pits and other types of defects on the end face during the polishing process. Hence, fiber end face inspection is a significant process for fiber manufactures when analysing the performance of a fiber. In order to identify the defects present on the fiber end face, a novel model is presented in this paper. Our model combined filtering methods to enhance the contrast of the images so scratches can be successfully detected. However, because the photos have been taken with different gains and exposures, they can not be processed with standard image processing techniques. We developed a method to analyse the defects intensity that could be located under different gains and exposures. We established that the images taken with the high gains and exposures performed well for optical fiber defect recognition.