The Application Of Image Segmentation In Detecting Metallographic Structure

This paper proposed an algorithm to segment the crystal structure of metallographic images using the methods of digital image processing. Because the traditional watershed transform usually result in over-segmentation, we improved the algorithm in the following steps. In the first stage, we transform the original image to binary image using double-threshold binaryzation, which is gray threshold and area threshold. In this method will it both preserve more effective information of crystal structure and suppress the information of noise. Then we modify the binary image by ridge detection, which will outline the main contour of the crystal structure. In the second stage, we segment the binary image which is gained above by iterative watershed transform. The"iterative"means that every time after the watershed transform, we find out the pseudo segmentation block and repartition it into other segmentation block, then do the next watershed transform segmentation until there is no pseudo segmentation block or reach the maximum iterative times the algorithm end. The algorithm of iterative watershed transform is implemented in Matlab7.0, the segmented images reflect most information of the crystal structure. The result is satisfactory.In this paper, firstly we make an introduction of Digital Image Processing, followed by an overview of traditional methods of image segmentation and its development. Then we give a detailed introduction of the segmentation algorithm of iterative watershed transform and some image processing methods related. At last, obtain the segmentation result.