Edge Information Extraction of Overlapping Fiber Optical Microscope Imaging Based on Modified Watershed Algorithm.

In the microscopic fields of materials, biology, etc., the research based on the images under the optical microscope is an important part in experiments. Because of the specific characteristics along with various materials, the automatic identification, calculation and statistical techniques of particles in the microscope imaging are facing barriers. Though a number of methods have been invented for edge segmentation, the watershed algorithm tends to display good performance in image feature analysis. However, its result of processing the fiber edge images under the microscope does not meet the requirements based on the experiments. This paper aims to propose an improved watershed algorithm in order to better solve the problem of over-segmentation and insufficient segmentation in the field of overlapping optical fiber images. The morphological algorithm is essentially processed as a pretreatment in the beginning. Then after the reinforcement, the OTSU algorithm is demanded as a more accurate binarization process in order to better distinguish the images. On this basis, the distance between the feature pixels is supposed to be calculated aiming to reconstruct the pixel value corresponding to each point. Experimental results are compared with the version obtained by the foreground background marking watershed algorithm.